Helium 10

10-character Amazon ASIN or Amazon product URL containing an ASIN; JSON key `asin_or_url`. Required and must not be blank. If the user has not provided an ASIN or product URL, ask them to supply it—do not infer, search for, or invent an ASIN.

Keywords to start tracking for this product; JSON key `keywords`. No size limit. Duplicates within the list, and keywords already tracked on this product, are merged or skipped automatically — you do not need to de-duplicate first. Output from get_keywords_new_suggestion can be passed straight through.

Amazon marketplace short code (e.g. `US`, `UK`, `DE`, `JP`); JSON key `marketplace`. REQUIRED for this write operation and has no default — every tracked product belongs to exactly one marketplace, so it cannot be omitted or inferred. Must be one of the uppercase marketplace codes exposed in this schema; lowercase or mixed case is rejected.

Which SHAPE to give the tracked product, chosen when tracking starts; JSON key `track_variations`. REQUIRED and has no default — if the user has not said which they want, ASK THEM before calling this tool; do not guess. true = the PARENT ASIN is tracked and its variations roll up into one entry. false = only the exact ASIN supplied is tracked. This is NOT an action that can be applied to a product that is already tracked: a tracked product's shape is fixed, and no tool (nor the Helium 10 web app) can change it afterwards. So do not send true to 'add variation tracking' to something already tracked — with an already-tracked parent that would only append your keywords to that parent and leave its shape untouched. To start tracking a specific variation of a product you already track, call this tool again with that variation's own child ASIN and `track_variations` false; it becomes a separate tracked product.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Whether to strip special characters from the keywords before tracking; JSON key `special_chars`. Defaults to true, matching the Helium 10 web 'add product' flow. Set to false only when the user explicitly wants punctuation preserved.

Explicit list of candidate phrases. Max 200 per call — split larger lists.

Marketplace code: US, UK, DE, JP, etc.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

REQUIRED. Marketplace short code (case-sensitive). Confirm the marketplace with the user once, before the first Listing Builder call of the session; every later Listing Builder tool reuses that same value — do NOT ask the user for it again once it is set. The output language for generated attributes is derived server-side from this value (e.g. 'JP' → Japanese). Supported values: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AU, JP, BE, BR.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Mixable with image_urls; combined at most 5. Each element is an opaque upload reference returned by request_image_upload_url AFTER the file was uploaded (channel 1 curl PUT, or channel 2 browser upload page). This is the REQUIRED path for local files.

Mixable with image_refs; combined at most 5. Each element MUST match EXACTLY ONE of: (a) HTTPS URL ending in .jpg/.jpeg/.png whose host resolves to a public IP (http and internal / private / link-local hosts are rejected as SSRF); (b) data URL 'data:image/<jpg|jpeg|png>;base64,<BASE64>' with a valid JPEG (FF D8 FF) or PNG (89 50 4E 47 0D 0A 1A 0A) magic header. Bare base64 (no 'data:' prefix), .webp, .gif are REJECTED. Catalog images from `retrieve_listing_by_asin` arrive here via `images[].url`. Do NOT base64 a local file into a data URL — it wastes conversation tokens and may be rejected; use request_image_upload_url + image_refs instead.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Request parameters for analyzing an Amazon (ASIN)product's performance

Request parameters for BSR chart query

Request parameters for review chart query

ASIN to check. Exactly 10 alphanumeric characters; uppercase and lowercase accepted. Server normalizes to uppercase for processing and echo.

Keywords to check for this ASIN. Prefer unique values. Duplicates are accepted and deduplicated server-side. One batch per call; Agent splits larger sets. Max 50.

Amazon marketplace code. Must be a supported uppercase enum value per platform MCP spec (e.g. US). Lowercase or mixed case is rejected.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional Seller ID. When provided, Storefront Index is included. Length must be 10-21 characters.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

1 to 10 competitor ASINs (must NOT include main_asin). Each ASIN is 10-char alphanumeric.

Primary ASIN under analysis (the user's product), 10-char alphanumeric (e.g. B09F9C3KFS).

Marketplace code: US, UK, DE, JP, etc. Must be in the Cerebro allowlist.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Opaque pagination cursor from a previous response's pagination.next_cursor. Omit for the first page. Pass it back verbatim with the same query arguments (limit may change between pages); never construct or edit it. Cursors are reusable but expire 30 minutes after being issued (pagination.cursor_expires_at) and are bound to the query that issued them.

Optional. Default false: the search covers each ASIN's parent and all child variations (main and every competitor). Pass true to cover only the exact ASINs given.

File format when results are delivered as a download (ignored for inline delivery). 'json' (default): gzip-compressed JSON, best for programmatic analysis. 'csv': UTF-8 CSV that opens directly in Excel/Sheets - prefer it when the user wants a spreadsheet/report file. Columns are the union of row fields; nested values stay JSON-encoded.

Number of rows to return, 1-10000 (default 150); out-of-range values are rejected. Up to 1000 rows come back inline; ask for more and the window is delivered as a file instead (data.export.download_url), so size `limit` to what is actually wanted. A page can still come back shorter than `limit` when the rows are unusually large (the ~1MB inline budget); pagination.truncated says so and `cursor` continues from there. When pagination.has_more is true, fetch the next batch with `cursor`. Do NOT probe with a small limit to learn the dataset size — every page carries the same total_count/has_more metadata, so ask for the rows you actually need and read the metadata of that response.

How results are delivered. 'auto' (default): rows are returned inline when they fit the response budget, otherwise they are exported to a file and data.export.download_url is returned instead — so simply request the amount of data you need via `limit`. 'inline': always inline, oversized pages are truncated and continue via cursor. 'download': always a downloadable file, even for small results; the file contains exactly the window selected by `limit`/`cursor`, so size `limit` to what is wanted (the tool's max for the whole dataset). Use 'download' when the user wants a file or when you need many rows for analysis (fetch the link in your execution environment instead of paging through the conversation).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

1 to 10 competitor Walmart item IDs (must NOT include main_item_id). Each is all digits. De-duplicated server-side; the resulting count is total_competitors, the CPS coverage denominator.

Primary Walmart item under analysis (the user's product), all digits (e.g. 10450114). It is the numeric segment at the end of a walmart.com product URL. Amazon ASINs are not accepted here - use compare_asin_keywords for Amazon.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Opaque pagination cursor from a previous response's pagination.next_cursor. Omit for the first page. Pass it back verbatim with the same query arguments (limit may change between pages); never construct or edit it. Cursors are reusable but expire 30 minutes after being issued (pagination.cursor_expires_at) and are bound to the query that issued them.

Optional. Default false: the search aggregates each item and all its variants (main and every competitor). Pass true to cover only the exact item IDs given.

File format when results are delivered as a download (ignored for inline delivery). 'json' (default): gzip-compressed JSON, best for programmatic analysis. 'csv': UTF-8 CSV that opens directly in Excel/Sheets - prefer it when the user wants a spreadsheet/report file. Columns are the union of row fields; nested values stay JSON-encoded.

Number of rows to return, 1-10000 (default 150); out-of-range values are rejected. Up to 1000 rows come back inline; ask for more and the window is delivered as a file instead (data.export.download_url), so size `limit` to what is actually wanted. A page can still come back shorter than `limit` when the rows are unusually large (the ~1MB inline budget); pagination.truncated says so and `cursor` continues from there. When pagination.has_more is true, fetch the next batch with `cursor`. Do NOT probe with a small limit to learn the dataset size — every page carries the same total_count/has_more metadata, so ask for the rows you actually need and read the metadata of that response.

How results are delivered. 'auto' (default): rows are returned inline when they fit the response budget, otherwise they are exported to a file and data.export.download_url is returned instead — so simply request the amount of data you need via `limit`. 'inline': always inline, oversized pages are truncated and continue via cursor. 'download': always a downloadable file, even for small results; the file contains exactly the window selected by `limit`/`cursor`, so size `limit` to what is wanted (the tool's max for the whole dataset). Use 'download' when the user wants a file or when you need many rows for analysis (fetch the link in your execution environment instead of paging through the conversation).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Up to 10 competitor ASINs (each 10 characters). Required: ``compare_listings`` only returns rows in multi-ASIN mode. Empty / blank entries are dropped before the upstream call.

Alias for ``main_asin`` (accepted because callers often send ``asin``). Prefer ``main_asin``; when both are supplied ``main_asin`` wins. Leave unset if you already pass ``main_asin``.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Main Amazon ASIN (10 characters). Always emitted as the first row of ``asins`` with ``is_main: true``. Required — the parameter is named ``main_asin`` (not ``asin``); you may also pass it via the ``asin`` alias.

Optional marketplace short code where the ASINs are listed. Supported: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AE, BR, AU. Defaults to US when omitted. Case-insensitive; whitespace is trimmed.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Request parameters for analyzing Helium 10 Competitor Listing

Setting amount, must be > 0 (no upper bound). For a multi-SKU product-level setting this is the TOTAL, split across the attached SKUs when expanded.

Human-readable name for the setting. Non-empty. Must be UNIQUE within the H10 account (case-sensitive); a duplicate name fails (not idempotent).

ISO 4217 currency code (e.g. USD). REQUIRED. For marketplace/product level it MUST equal that marketplace's currency; for h10_account/seller level any supported code is accepted.

IANA timezone (e.g. America/Los_Angeles). REQUIRED. Must be in the Amazon-report timezone allowlist. For marketplace/product level it MUST equal that marketplace's canonical report timezone; for h10_account/seller level any allowlisted value is accepted. SCOPE OF EFFECT — read before relying on this field: date_tz only fixes which calendar day the setting's created_at is stamped in. It is NOT stored on the setting and is NOT returned by the list tools, and it does NOT make the cost timezone-scoped: starts_at / ends_at are plain calendar dates, each expanded charge carries a plain occurrence_date, and P&L attributes that charge to that calendar day with no timezone conversion. Per-cost timezone billing is therefore not supported — to vary by region, create one setting per marketplace (each marketplace implies its own report timezone).

Recurrence: one_time | daily | weekly | monthly | quarterly | half_yearly | annually.

Effective start date, yyyy-MM-dd (no time component). Required.

When true and frequency is coarser than daily, the amount is spread evenly across the days in each period; false/omit = discrete period anchor points. IGNORED (treated as false) for one_time and daily frequencies.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Effective end date, yyyy-MM-dd. Omit or null = open-ended (recurring into the future). one_time may omit it. When both set, starts_at <= ends_at.

Short country code (US, CA, UK, DE, …) — NEVER an Amazon marketplace ID or domain. Requires seller_id. Present -> marketplace level or narrower.

Amazon seller account ID (must be authorised on the token). Present -> seller level or narrower. Omit for an h10_account-level setting.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Product-level SKUs to attach. Non-empty when provided; requires seller_id + marketplace (product level). Each element is {sku, asin?}. Omit for h10_account/seller/marketplace level.

Opaque COGS record IDs from list_amazon_cogs (1..100). Each must be a non-empty numeric id string; empty/blank/malformed -> INVALID_INPUT (whole request rejected). Duplicates are de-duplicated server-side. IDs outside token scope / unknown / already-deleted are silently ignored (deleted_count may be < ids sent).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Opaque Indirect Cost setting IDs from list_amazon_indirect_cost_settings (1..100). Each must be a non-empty numeric id string; empty / blank / malformed → INVALID_INPUT (whole request rejected). Duplicates are de-duplicated server-side. IDs outside token scope / unknown / already-deleted are silently ignored (deleted_count may be < ids sent).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

10-character Amazon ASIN or an Amazon product URL containing one; JSON key `asin_or_url`. Must be the ASIN as it appears in the user's tracking list — it is matched exactly, and a product tracked as a parent with all its variations is stored under the PARENT ASIN. If the user has not provided an ASIN or product URL, ask them to supply it—do not infer, search for, or invent an ASIN.

Keywords to stop tracking on this product; JSON key `keywords`. No size limit. Must match the tracking list exactly (case, spacing and punctuation included) — pass the strings returned by list_tracked_keywords verbatim. Sent as-is: repeats are reported back in duplicate_count and blank entries in failed_keywords, so every entry you send is accounted for. Do not pre-filter the list yourself.

Amazon marketplace short code (e.g. `US`, `UK`, `DE`, `JP`); JSON key `marketplace`. REQUIRED and has no default — every tracked product belongs to exactly one marketplace, so it cannot be omitted or inferred. Must be one of the uppercase marketplace codes exposed in this schema; lowercase or mixed case is rejected.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

10-character Amazon ASIN or an Amazon product URL containing one; JSON key `asin_or_url`. Must be the ASIN as it appears in the user's tracking list — it is matched exactly, and a product tracked as a parent with all its variations is stored under the PARENT ASIN. If the user has not provided an ASIN or product URL, ask them to supply it—do not infer, search for, or invent an ASIN.

Amazon marketplace short code (e.g. `US`, `UK`, `DE`, `JP`); JSON key `marketplace`. REQUIRED and has no default — every tracked product belongs to exactly one marketplace, so it cannot be omitted or inferred. Must be one of the uppercase marketplace codes exposed in this schema; lowercase or mixed case is rejected.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Batch of COGS upsert items (1..100). Overlay upsert located by seller_id + marketplace + sku (ID-free). JSON array order is preserved; error index is 0-based into this array.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

New setting amount, must be > 0. For a multi-SKU product setting this is the TOTAL, re-split across SKUs when charges regenerate.

Opaque setting id from list_amazon_indirect_cost_settings. Must belong to the token's H10 account; unknown / foreign id → error (own-only).

New or unchanged setting name. Must stay UNIQUE within the H10 account; renaming to an existing name → error.

Effective start date, yyyy-MM-dd (no time). Required.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Effective end date, yyyy-MM-dd. Omit or null = open-ended. When both set, starts_at <= ends_at.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Which way to switch monitoring. 'enable' starts collecting alerts for the listing; 'disable' stops it and permanently gives up the change events of the off-period. No default — state it explicitly.

Amazon ASIN to switch monitoring for, 10 alphanumeric characters.

Marketplace code, e.g. US.

Helium 10 seller identifier.

Seller SKU of the listing to switch. Required: one seller can list the same ASIN under several SKUs in a marketplace, each with its own monitoring switch. Take it from get_monitoring_status(asin).asin_rows[].seller_sku.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

One or more actions to execute against this platform + profileId. Each has a ``type`` (from get_ads_campaign_mgmt_platforms) and a ``request_body`` (shape from get_ads_campaign_mgmt_schema). Count must not exceed the platform's maxActionsPerRequest.

Platform key from get_ads_campaign_mgmt_platforms (e.g. amazon, walmart). Case-sensitive.

Target profile/account id (required). Resolve via get_ads_query_materials (level=profile for non-DSP; advertiser for amazon-dsp; vendor_account / seller_account for commerce). Wire field: ``profileId``.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Change type from get_ads_changelog_platforms; drives which filters are valid (see get_ads_changelog_schema).

Range start, DD-MM-YYYY (e.g. 30-07-2026).

Range end, DD-MM-YYYY (e.g. 31-07-2026).

Rows per page (1–500). Recommended: 100.

Platform key (amazon, walmart). Case-sensitive.

Target profile id from get_ads_query_materials(level=profile).

Optional campaign ids from get_ads_query_materials(level=campaign).

Optional; valid only when change_source is auto/semi_auto; value from schema.changeBy.byChangeSource[change_source].

Optional schema.changeSource value. 'all' = no filter -> OMIT. Only auto/semi_auto enable change_by.

Optional. AI values come from schema.changeSourceId.byChangeBy.ai.static (use directly). Rule ids (hex32) and schedule uuids are account data the caller already holds.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

1-based page number (default 1).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Sort order. Default '-changed_at' (newest first); only changed_at is sortable.

Optional keyword/PAT text. Only when schema.filters.target_name.available=true (bid / keyword / negative change types).

Level name (e.g. profile, campaign, target). Same value used in get_ads_query_schema.

Platform key (e.g. amazon, walmart). Same value used in get_ads_query_list and get_ads_query_schema.

Period-over-period comparison config. Default null (no comparison).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Group-by dimensions, as a list of OBJECTS — each {"field": <a dimension field name from get_ads_query_schema>} with optional "granularity" for date fields. NOT bare strings. Example (illustrative — use real fields from the schema): [{"field":"campaign_id"},{"field":"campaign_name"}]. Contrast measures, which are bare strings. Default null (no grouping; one aggregate row).

WHERE-clause conditions, AND-combined; each an object {"field","op","value"} with field + op from get_ads_query_schema (filters[].field / filters[].ops). For enum fields pass the enum's `value` verbatim (case-sensitive), not the `physical` form. There is no date parameter — filter by the schema's date field, e.g. op "between" with [start, end] (illustrative: {"field":"date","op":"between","value":["2026-06-01","2026-06-30"]}; confirm the field name and op in the schema). State filtering is supported — add a not_in/in/eq filter on the state field only when the request implies it; no exclude-by-default.

HAVING-clause conditions on aggregated measures, AND-combined. Default null.

Max rows to return. Default null (backend default).

Measure field names, as a list of BARE STRINGS taken from get_ads_query_schema measures[].field (pass the name only). NOT objects — the schema lists measures as {field,...} objects (a catalog of what's available), but here pass just the string. Example (illustrative — use real fields from the schema): ["impressions","clicks","spend"]. Contrast dimensions, which are objects. Default null (dimension-only query).

Row offset for pagination. Default null (no offset).

Query-level options. Default null -- when omitted, the ``options`` field is not sent on the wire at all. When you pass an object (even ``{}``), the per-field defaults apply: ``attribution_window="14d"`` and ``to_currency=""`` (empty string, which the backend interprets as 'no currency conversion'). Pass an explicit ``null`` on a sub-field to opt out of that default.

Sort order (the parameter is named order_by, NOT sort). List of objects {"field": <a field from get_ads_query_schema>, "direction": "asc|desc"}, e.g. [{"field":"spend","direction":"desc"}] (illustrative). Multi-key sort applies items in array order. Default null (backend-default ordering).

Tenant-isolation profile ids (string[]). Optional on the input surface; the upstream swagger marks ``profileIds`` as nullable and accepts an empty list as 'every profile the user has access to' (tenant isolation is still enforced via the inbound bearer token). Forward an explicit list when the user named a specific profile (resolve via get_ads_query_materials(level=profile); each returned row carries the profile id as a number, so stringify it before forwarding -- e.g. ``["100001", "100002"]``). When omitted, the wrapper emits ``profileIds: []`` on the wire, which the backend interprets as the user's full accessible set. Wire field: ``profileIds``.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

REQUIRED. Marketplace short code (case-sensitive). Confirm the marketplace with the user once, before the first Listing Builder call of the session; every later Listing Builder tool reuses that same value — do NOT ask the user for it again once it is set. Supported values: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AU, JP, BE, BR.

One or more keyword sources to fetch from. If the user has no preference, default to aba_converting_keywords + top_keywords. 1. `aba_converting_keywords` (default): ABA Converting Keywords — Amazon Brand Analytics terms that convert to sales in this niche. ⚠️ Returns NO keywords unless the user's Amazon store is connected/bound in Helium 10. 2. `sqp_sales_keywords`: SQP Sales Keywords — Search Query Performance terms driving the most purchases for the ASIN. ⚠️ Returns NO keywords unless the user's store is connected/bound in Helium 10 AND the main ASIN or a competitor ASIN belongs to that store. 3. `top_keywords` (default): Top Keywords — the highest-traffic keywords the ASIN and its competitors already rank for. 4. `opportunity_keywords`: Opportunity Keywords — high-traffic terms where the ASIN ranks weakly (ranking gaps to capture). 5. `historical_top_keywords`: Historical Top Keywords — top keywords aggregated over a longer historical window. 6. `top_performing_organic_keywords`: Top Performing Organic Keywords — the best organic-rank keywords for the ASIN / competitors. 7. `top_performing_sponsored_keywords`: Top Performing Sponsored Keywords — the best-performing sponsored (PPC) keywords.

Main ASIN (10-character alphanumeric Amazon Standard Identification Number). OPTIONAL — provide at least one of asin / competitor_asins.

Competitor ASINs (up to 9). OPTIONAL — provide at least one of asin / competitor_asins; supplying BOTH is strongly recommended. CPS / competitor-advantage sources and ranking metrics rely on competitor data, so include competitors whenever available.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Reserved for future use — accepted but IGNORED in this release.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

2D array of keyword allocations for each bullet point. Outer list length = number of bullets (5 to 10; recommend 5). Each inner list: default 3 keywords, max 6 keywords, semantically aligned. Derive from the user-approved allocation of the find_keywords_with_multi_source Keyword Bank.

REQUIRED. Marketplace short code (case-sensitive). Confirm the marketplace with the user once, before the first Listing Builder call of the session; every later Listing Builder tool reuses that same value — do NOT ask the user for it again once it is set. Supported values: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AU, JP, BE, BR.

Structured product attributes derived from `analyze_product_images` (the `analyzed_product` field). Pass the object back verbatim from `analyze_product_images` with any user-confirmed corrections applied; at minimum product_name should be set for best results.

The FULL find_keywords_with_multi_source Keyword Bank — pass EVERY row, not just the allocated ones, PLUS any extra keywords the user added during the allocation review. The backend subtracts the keywords you allocated to title / bullet / description slots and feeds the remainder to the AI as optional keywords to weave into the copy, so passing the full bank improves the generated listing. The `keyword` field maps to `phrase`; other metric fields (search_volume, cps, starred, suggested, priority_score, keyword_sources) are all optional — pass whatever the Keyword Bank row carries. Unrelated columns (cerebro_iq_score, cpr, etc.) are dropped.

(Optional) Competitor ASIN whose style/tone to mimic, collected during the analyze_product_images review.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Keywords for the description slot (default 3, max 6), from the user-approved allocation.

Maximum character length per bullet point (50-700). Defaults to 250.

Minimum character length per bullet point (50-700). Defaults to max(50, max_bullet_len - 40) (i.e. 210 when max_bullet_len is 250). Set explicitly to override.

request_id from `analyze_product_images` (links records for audit).

(Optional) Up to 2 ASINs whose Alexa for Shopping Q&A content (formerly Rufus Q&A — Amazon rebranded Rufus to Alexa for Shopping in 2026) should be woven into the listing. Collected during the analyze_product_images review. Parameter name kept as `rufus_asins` for backward compatibility; will be renamed to `alexa_qa_asins` in a future release.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Keywords for the title slot (default 3, max 6), from the user-approved allocation. Keep the title concise (<=75 chars); keywords that don't fit belong in item_highlight, not a longer title.

Helium 10 backend API path or full URL. Must point to an allow-listed Helium 10 / Pacvue host (h10api.pacvue.com, h10api-test.pacvue.com, members.helium10.com).

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

Optional. SQL IN filter by ASIN list (array of strings, e.g. ['B001234567']). Default null (all ASINs).

Optional. Only 11 metrics support sub-component breakdown: sales, units_sold, refund, advertising_cost, amazon_fees, reimbursements, cost_of_goods, indirect_cost, gross_revenue, sessions, units_returned. The other 10 (promo, estimated_payout, gross_profit, expenses, net_profit, margin, roi, tacos, ads_acos, refunds_percentage) are derived metrics — use summary tools for those. Default null = all 11.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Optional, defaults to USD.

Timezone used to interpret calendar dates. Must be one of the Amazon marketplace report timezones (not any IANA timezone). Examples: America/Los_Angeles, Europe/Paris, Europe/London. Italy/Germany/Spain use Europe/Paris (not Europe/Rome or Europe/Berlin). Do not pass a city name, abbreviation, or UTC offset. Optional; when omitted, defaults to the queried marketplace's report timezone (America/Los_Angeles when no single marketplace is set).

Optional marketplace filter as an array of region short codes ONLY (e.g. ['US', 'CA', 'MX']); Amazon SP-API marketplace IDs (e.g. 'ATVPDKIKX0DER'), seller IDs, and domains are rejected. Allowed: AE, AU, BE, BR, CA, CN, DE, EG, ES, FR, IE, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US, ZA. Default null (all authorised marketplaces).

Optional. SQL IN filter on seller account IDs (array of strings). Profits intersects with sellers from Pacvue token; no intersection → empty result. Default null (all authorised sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by seller SKU list (array of strings). Default null (all SKUs).

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

Time-bucket size for series tools: day | week | month. Use the exact parameter name granularity; do not use scope, interval, period, or bucket as input aliases. Week/month align to calendar boundaries (week starts Monday; month starts on the 1st); first/last buckets may be partial. The first bucket key may precede current_date_from (e.g. from 2026-04-01 at week granularity → first key 2026-03-30). Partial buckets include only data inside the requested date range. Series responses no longer include the scope or offset fields (removed in HP-1495); use the request granularity you sent.

Optional. SQL IN filter by ASIN list (array of strings, e.g. ['B001234567']). Default null (all ASINs).

Optional. Only 11 metrics support sub-component breakdown: sales, units_sold, refund, advertising_cost, amazon_fees, reimbursements, cost_of_goods, indirect_cost, gross_revenue, sessions, units_returned. The other 10 (promo, estimated_payout, gross_profit, expenses, net_profit, margin, roi, tacos, ads_acos, refunds_percentage) are derived metrics — use summary tools for those. Default null = all 11.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Optional, defaults to USD.

Timezone used to interpret calendar dates. Must be one of the Amazon marketplace report timezones (not any IANA timezone). Examples: America/Los_Angeles, Europe/Paris, Europe/London. Italy/Germany/Spain use Europe/Paris (not Europe/Rome or Europe/Berlin). Do not pass a city name, abbreviation, or UTC offset. Optional; when omitted, defaults to the queried marketplace's report timezone (America/Los_Angeles when no single marketplace is set).

Optional marketplace filter as an array of region short codes ONLY (e.g. ['US', 'CA', 'MX']); Amazon SP-API marketplace IDs (e.g. 'ATVPDKIKX0DER'), seller IDs, and domains are rejected. Allowed: AE, AU, BE, BR, CA, CN, DE, EG, ES, FR, IE, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US, ZA. Default null (all authorised marketplaces).

Optional. SQL IN filter on seller account IDs (array of strings). Profits intersects with sellers from Pacvue token; no intersection → empty result. Default null (all authorised sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by seller SKU list (array of strings). Default null (all SKUs).

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

Optional. SQL IN filter by ASIN list (array of strings, e.g. ['B001234567']). Default null (all ASINs).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Optional, defaults to USD.

Timezone used to interpret calendar dates. Must be one of the Amazon marketplace report timezones (not any IANA timezone). Examples: America/Los_Angeles, Europe/Paris, Europe/London. Italy/Germany/Spain use Europe/Paris (not Europe/Rome or Europe/Berlin). Do not pass a city name, abbreviation, or UTC offset. Optional; when omitted, defaults to the queried marketplace's report timezone (America/Los_Angeles when no single marketplace is set).

Optional marketplace filter as an array of region short codes ONLY (e.g. ['US', 'CA', 'MX']); Amazon SP-API marketplace IDs (e.g. 'ATVPDKIKX0DER'), seller IDs, and domains are rejected. Allowed: AE, AU, BE, BR, CA, CN, DE, EG, ES, FR, IE, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US, ZA. Default null (all authorised marketplaces).

Optional. SQL IN filter on seller account IDs (array of strings). Profits intersects with sellers from Pacvue token; no intersection → empty result. Default null (all authorised sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by seller SKU list (array of strings). Default null (all SKUs).

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

Time-bucket size for series tools: day | week | month. Use the exact parameter name granularity; do not use scope, interval, period, or bucket as input aliases. Week/month align to calendar boundaries (week starts Monday; month starts on the 1st); first/last buckets may be partial. The first bucket key may precede current_date_from (e.g. from 2026-04-01 at week granularity → first key 2026-03-30). Partial buckets include only data inside the requested date range. Series responses no longer include the scope or offset fields (removed in HP-1495); use the request granularity you sent.

Optional. SQL IN filter by ASIN list (array of strings, e.g. ['B001234567']). Default null (all ASINs).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Optional, defaults to USD.

Timezone used to interpret calendar dates. Must be one of the Amazon marketplace report timezones (not any IANA timezone). Examples: America/Los_Angeles, Europe/Paris, Europe/London. Italy/Germany/Spain use Europe/Paris (not Europe/Rome or Europe/Berlin). Do not pass a city name, abbreviation, or UTC offset. Optional; when omitted, defaults to the queried marketplace's report timezone (America/Los_Angeles when no single marketplace is set).

Optional marketplace filter as an array of region short codes ONLY (e.g. ['US', 'CA', 'MX']); Amazon SP-API marketplace IDs (e.g. 'ATVPDKIKX0DER'), seller IDs, and domains are rejected. Allowed: AE, AU, BE, BR, CA, CN, DE, EG, ES, FR, IE, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US, ZA. Default null (all authorised marketplaces).

Optional. SQL IN filter on seller account IDs (array of strings). Profits intersects with sellers from Pacvue token; no intersection → empty result. Default null (all authorised sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by seller SKU list (array of strings). Default null (all SKUs).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Action type from get_ads_campaign_mgmt_platforms (actions[].actionType), e.g. campaign_create, keyword_update. Case-sensitive. Wire param: ``actionType``.

Platform key from get_ads_campaign_mgmt_platforms (e.g. amazon, walmart). Case-sensitive.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional. Restrict the result to one platform. Must be a plain string: 'amazon' or 'walmart'. Do not pass an array or object.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Change type from get_ads_changelog_platforms (changeTypes[].changeType), e.g. bid_changed. Case-sensitive.

Platform key from get_ads_changelog_platforms (amazon, walmart). Required. Case-sensitive.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

A platform key returned by get_ads_query_platforms (e.g. amazon, walmart). Case-sensitive.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Group-by dimensions (REQUIRED, non-empty), as a list of OBJECTS — each {"field": <a field name from get_ads_query_schema dimensions[].field>}. NOT bare strings. Example (illustrative — use real fields from the schema): [{"field":"profile_id"},{"field":"profile_name"}]. Typically the entity id and display name for the level. No granularity here — material listing never time-buckets.

Material level name (e.g. profile, campaign, adgroup, target). Same value used in get_ads_query_schema; names come from get_ads_query_list.levels[].name.

Platform key (e.g. amazon, walmart). Same value used in get_ads_query_list and get_ads_query_schema.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

WHERE-clause conditions, AND-combined. Use this to narrow the entity list (e.g. filter campaigns by name substring, by state, by adgroup_id, ...). By UI convention, append a state-like 'not_in' filter to exclude delete-like enum values (archived / deleted / removed / ...) unless the user explicitly asks for them. Read enum values from the schema and pass the `value` verbatim (case-sensitive), not the `physical` form. Default null.

Max rows to return. Default null (backend default).

Row offset for pagination. Default null (no offset).

Tenant-isolation profile ids (string[]). Optional on the input surface (the upstream swagger marks ``profileIds`` as nullable). Pass an explicit list to restrict the materials to those profiles; leave empty / null to list materials across every profile the user has access to (the upstream accepts the empty list and falls back to the bearer token's accessible set, tenant isolation stays enforced). Each material id returned by this tool is a number; stringify it before forwarding elsewhere (e.g. ``["100001", "100002"]``). Wire field: ``profileIds``.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

A valid level name from get_ads_query_list (e.g. profile, campaign, adgroup, target). Case-sensitive.

Platform key (same value used in get_ads_query_list).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

The rule_id (source=live) or draft_id (source=draft) to fetch. For source=live this is a 32-character lowercase hex rule_id (e.g. faa679c9b78e4b9fb8b452326f24dd00); for source=draft it is a decimal draft_id (e.g. 2089352922874187778).

'live' to fetch a committed rule by rule_id; 'draft' to fetch a draft by draft_id.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional platform hint (e.g. amazon).

Target profile_id. Optional for source=draft; for source=live it scopes the tenant -- omit it only if you want the server to auto-resolve your single accessible profile.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Platform key from get_ads_rule_platforms (e.g. amazon). Case-sensitive.

Rule type from get_ads_rule_platforms (e.g. BID). Case-sensitive.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Range start, ISO YYYY-MM-DD (e.g. 2026-08-01). Note this differs from execute_ads_changelog_query on the same server, which takes DD-MM-YYYY.

Range end, ISO YYYY-MM-DD (e.g. 2026-08-19).

Target Amazon profile id from get_ads_query_materials(level=profile).

Optional numeric Amazon campaign ids. Omit to cover every campaign of the profile. An id that does not exist in this profile is not an error -- it simply does not appear in rows, so reconcile the result against what you asked for.

Optional campaign state filter. Only enabled, paused and archived are accepted; other states are rejected rather than silently ignored.

Optional campaign type filter. Only sponsoredProducts, sponsoredBrands, sponsoredDisplay and sponsoredTV are accepted; other values are rejected rather than silently ignored.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Required. Numeric Amazon campaign ids to read, at most 500 per call. Amazon's budget-usage API has no 'all campaigns' mode, so resolve the ids first (e.g. get_ads_query_materials at campaign level, or the ids from a prior query).

Target Amazon profile id from get_ads_query_materials(level=profile).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Required. ASIN to query, 10-char alphanumeric (e.g. B01LR5S6HK).

Required. Amazon marketplace code. Allowed values: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AU, JP, AE, BR, SA.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional window end (inclusive), format yyyy-MM-dd (UTC day), no later than today. Pass it together with start_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional window start (inclusive), format yyyy-MM-dd (UTC day). Pass it together with end_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

Required. ASIN to query, 10-char alphanumeric (e.g. B01LR5S6HK).

Required. Amazon marketplace code. Allowed values: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AU, JP, AE, BR, SA.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional window end (inclusive), format yyyy-MM-dd (UTC day), no later than today. Pass it together with start_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional window start (inclusive), format yyyy-MM-dd (UTC day). Pass it together with end_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

Amazon ASIN to analyze (child, parent, or standalone). Exactly ONE ASIN per call, exactly 10 alphanumeric characters. More than one ASIN is rejected with INVALID_INPUT.INVALID_ASIN_FORMAT — input is never silently truncated.

Amazon marketplace code. Two-letter UPPERCASE code; supported values: US, UK, FR, IT, DE, ES, JP. Lowercase or unsupported codes are rejected with INVALID_INPUT.INVALID_MARKETPLACE.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Required. ASIN to query, 10-char alphanumeric (e.g. B01LR5S6HK).

Required. Amazon marketplace code. Allowed values: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AU, JP, AE, BR, SA.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional window end (inclusive), format yyyy-MM-dd (UTC day), no later than today. Pass it together with start_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional window start (inclusive), format yyyy-MM-dd (UTC day). Pass it together with end_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

Required. Marketplace-specific product ID to query — the ASIN on Amazon, 10-char alphanumeric (e.g. B01LR5S6HK).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Opaque pagination cursor from a previous response's pagination.next_cursor. Omit for the first page. Pass it back verbatim with the same query arguments (limit may change between pages); never construct or edit it. Cursors are reusable but expire 30 minutes after being issued (pagination.cursor_expires_at) and are bound to the query that issued them.

Optional window end (inclusive), format yyyy-MM-dd (UTC day), no later than today. Pass it together with start_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

File format when results are delivered as a download (ignored for inline delivery). 'json' (default): gzip-compressed JSON, best for programmatic analysis. 'csv': UTF-8 CSV that opens directly in Excel/Sheets - prefer it when the user wants a spreadsheet/report file. Columns are the union of row fields; nested values stay JSON-encoded.

Optional projection: return only these row fields, to save tokens when you need a few columns. Unknown names get a self-healing error listing the available fields. Overrides detail_level when both are set. Omit for the full field set.

Number of rows to return. Defaults to 100, max 10000; out-of-range values are clamped rather than rejected. Rows are large (~90B per row), so request only what the user needs — up to 500 rows per page come back inline (slightly larger asks are truncated and continue via `cursor`); requests far above the inline budget (>1000 rows or very large payloads) are delivered as a download link (response_format=auto). When pagination.has_more is true, fetch the next batch with `cursor`. Do NOT probe with limit=1 to learn the dataset size — every page carries the same total_count/has_more metadata, so ask for the rows you actually need and read the metadata of that response.

Amazon marketplace code, default US. Allowed values: US, CA, MX, BR, DE, ES, FR, IT, UK, NL, TR, IN, AE, EG, SA, SE, PL, JP, AU, SG, CN, BE, ZA, IE.

How results are delivered. 'auto' (default): rows are returned inline when they fit the response budget, otherwise they are exported to a file and data.export.download_url is returned instead — so simply request the amount of data you need via `limit`. 'inline': always inline, oversized pages are truncated and continue via cursor. 'download': always a downloadable file, even for small results; the file contains exactly the window selected by `limit`/`cursor`, so size `limit` to what is wanted (the tool's max for the whole dataset). Use 'download' when the user wants a file or when you need many rows for analysis (fetch the link in your execution environment instead of paging through the conversation).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional window start (inclusive), format yyyy-MM-dd (UTC day). Pass it together with end_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional marketplace filter (US, UK, DE, JP, CA, FR, IT, ES). When set, only seller rows whose marketplaces[] contains this value are returned. Does not trim marketplaces[] on each row.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Helium 10 seller identifier.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional marketplace code. Omit for an all-marketplaces overview; pass US/UK/DE/JP/etc. for single-market detail.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Confirmed numeric Amazon browse node id (from search_browse_nodes after user confirmation), e.g. "3733721". Exactly ONE browse node id per call; multiple ids, category names, or any other non-numeric value are rejected with INVALID_INPUT.INVALID_BROWSE_NODE — input is never silently truncated.

Amazon marketplace code. Two-letter UPPERCASE code; supported values: US, UK, FR, IT, DE, ES, JP. Lowercase or unsupported codes are rejected with INVALID_INPUT.INVALID_MARKETPLACE.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Marketplace code: US, UK, DE, JP, etc.

Helium 10 seller identifier.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Aggregation window: '24h' (default) or '7d'.

Alias for ``main_asin`` (accepted because callers often send ``asin``). Prefer ``main_asin``; when both are supplied ``main_asin`` wins. Leave unset if you already pass ``main_asin``.

Optional list of up to 10 competitor ASINs (each 10 characters). Omit (or pass an empty list) to let upstream auto-discover competitors and return their group average only (no per-competitor rows); supply explicit ASINs to instead receive per-ASIN sales blocks plus the cross-ASIN seed average.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Main Amazon ASIN to benchmark (10 characters). Always the anchor of the competitor comparison. Required — the parameter is named ``main_asin`` (not ``asin``); you may also pass it via the ``asin`` alias.

Optional marketplace short code where the ASIN is listed. Supported: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AE, BR, AU. Defaults to US when omitted. Case-insensitive; whitespace is trimmed.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by ASIN list (array of strings, e.g. ['B001234567']). Default null (all ASINs).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional. Maximum inclusive health.days_of_supply. The bound itself must be a non-negative integer, but it is compared against a fractional value: only rows where days_of_supply <= days_of_supply_max are returned, so days_of_supply_max=14 keeps 14.0 and 13.9 but drops 14.5. Rows with null days_of_supply are excluded when this or days_of_supply_min is set. May be combined with days_of_supply_min.

Optional. Minimum inclusive health.days_of_supply. The bound itself must be a non-negative integer, but it is compared against a fractional value: only rows where days_of_supply >= days_of_supply_min are returned, so days_of_supply_min=14 keeps 14.0 and 14.5 but drops 13.9. Rows with null days_of_supply are excluded when this or days_of_supply_max is set. May be combined with days_of_supply_max.

Optional. Inventory age bucket presence filter. Returns only rows that HAVE units (count > 0) in the selected bucket. Allowed values: less_31_days, 31_to_180_days, over_180_days. At most one value (mutually exclusive); omit for no age-stage filter. This is a presence filter, not exclusivity: a matching row still reports all three inv_age_* counts, so a SKU matching over_180_days may also have younger inventory.

Optional marketplace filter as an array of region short codes ONLY (e.g. ['US', 'CA', 'MX']); Amazon SP-API marketplace IDs (e.g. 'ATVPDKIKX0DER'), seller IDs, and domains are rejected. Allowed: AE, AU, BE, BR, CA, CN, DE, EG, ES, FR, IE, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US, ZA. Default null (all authorised marketplaces).

Optional. Page number (integer), starts from 1. Default 1.

Optional. Rows per page (integer). Min 5, max 1000. Default 50.

Optional. SQL IN filter on seller account IDs (array of strings). Profits intersects with sellers from Pacvue token; no intersection → empty result. Default null (all authorised sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by seller SKU list (array of strings). Default null (all SKUs).

Optional. SQL IN filter by ASIN list (array of strings, e.g. ['B001234567']). Default null (all ASINs).

Optional. Maximum inclusive inventory.available quantity (non-negative integer). When set, only rows where available <= available_max are returned. May be combined with available_min.

Optional. Minimum inclusive inventory.available quantity (non-negative integer). When set, only rows where available >= available_min are returned. May be combined with available_max.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional fulfillment channel filter: FBA | FBM. Omit to query both FBA and FBM inventory.

Optional marketplace filter as an array of region short codes ONLY (e.g. ['US', 'CA', 'MX']); Amazon SP-API marketplace IDs (e.g. 'ATVPDKIKX0DER'), seller IDs, and domains are rejected. Allowed: AE, AU, BE, BR, CA, CN, DE, EG, ES, FR, IE, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US, ZA. Default null (all authorised marketplaces).

Optional. Page number (integer), starts from 1. Default 1.

Optional. Rows per page (integer). Min 5, max 1000. Default 50.

Optional. SQL IN filter on seller account IDs (array of strings). Profits intersects with sellers from Pacvue token; no intersection → empty result. Default null (all authorised sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by seller SKU list (array of strings). Default null (all SKUs).

Optional. Metric to sort product rows by, using each row's inventory.available value. If omitted, server default order applies (unspecified, stable). Allowed values: available.

Optional. Sort direction for sort_by. 'desc' = highest value first (default); 'asc' = lowest value first. Ignored if sort_by is not provided.

1-300 keywords to evaluate. The cap matches find_keywords_with_multi_source's return cap, so its rows can be passed through whole. If you have more, pass the highest-priority_score subset.

REQUIRED. Your product's ASIN (10-character alphanumeric). Always first in `asins` with is_mine=true.

REQUIRED. Marketplace short code (case-sensitive). No default — matching is language-specific. Confirm the marketplace with the user once, before the first Listing Builder call of the session; every later Listing Builder tool reuses that same value — do NOT ask the user for it again once it is set. Supported values: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AU, JP, BE, BR.

0-10 competitor ASINs. Empty means single-listing keyword performance only (metrics.cps is then null). main_asin and duplicates are dropped silently and reported in skipped_asins.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Unpublished copy to score for the main ASIN instead of its published listing. Field-by-field override.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Single Amazon ASIN, 10-char alphanumeric (e.g. B01LR5S6HK).

Marketplace code: US, UK, DE, JP, etc. Must be in the Cerebro allowlist.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Opaque pagination cursor from a previous response's pagination.next_cursor. Omit for the first page. Pass it back verbatim with the same query arguments (limit may change between pages); never construct or edit it. Cursors are reusable but expire 30 minutes after being issued (pagination.cursor_expires_at) and are bound to the query that issued them.

Optional. Default false: the search aggregates the ASIN's parent and all child variations. Pass true to cover only the exact ASIN given.

File format when results are delivered as a download (ignored for inline delivery). 'json' (default): gzip-compressed JSON, best for programmatic analysis. 'csv': UTF-8 CSV that opens directly in Excel/Sheets - prefer it when the user wants a spreadsheet/report file. Columns are the union of row fields; nested values stay JSON-encoded.

Number of rows to return, 1-10000 (default 150); out-of-range values are rejected. Up to 1000 rows come back inline; ask for more and the window is delivered as a file instead (data.export.download_url), so size `limit` to what is actually wanted. A page can still come back shorter than `limit` when the rows are unusually large (the ~1MB inline budget); pagination.truncated says so and `cursor` continues from there. When pagination.has_more is true, fetch the next batch with `cursor`. Do NOT probe with a small limit to learn the dataset size — every page carries the same total_count/has_more metadata, so ask for the rows you actually need and read the metadata of that response.

How results are delivered. 'auto' (default): rows are returned inline when they fit the response budget, otherwise they are exported to a file and data.export.download_url is returned instead — so simply request the amount of data you need via `limit`. 'inline': always inline, oversized pages are truncated and continue via cursor. 'download': always a downloadable file, even for small results; the file contains exactly the window selected by `limit`/`cursor`, so size `limit` to what is wanted (the tool's max for the whole dataset). Use 'download' when the user wants a file or when you need many rows for analysis (fetch the link in your execution environment instead of paging through the conversation).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional historical month YYYY-MM (e.g. 2025-12). Omit for current data. One month per request; must fall within the 24-month lookback ending at the current UTC month.

Single Walmart item ID, all digits (e.g. 10450114). It is the numeric segment at the end of a walmart.com product URL. Amazon ASINs are not accepted here - use get_keywords_by_asin for Amazon.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Opaque pagination cursor from a previous response's pagination.next_cursor. Omit for the first page. Pass it back verbatim with the same query arguments (limit may change between pages); never construct or edit it. Cursors are reusable but expire 30 minutes after being issued (pagination.cursor_expires_at) and are bound to the query that issued them.

Optional. Default false: the search aggregates the item and all its variants. Pass true to cover only the exact item ID given.

File format when results are delivered as a download (ignored for inline delivery). 'json' (default): gzip-compressed JSON, best for programmatic analysis. 'csv': UTF-8 CSV that opens directly in Excel/Sheets - prefer it when the user wants a spreadsheet/report file. Columns are the union of row fields; nested values stay JSON-encoded.

Number of rows to return, 1-10000 (default 150); out-of-range values are rejected. Up to 1000 rows come back inline; ask for more and the window is delivered as a file instead (data.export.download_url), so size `limit` to what is actually wanted. A page can still come back shorter than `limit` when the rows are unusually large (the ~1MB inline budget); pagination.truncated says so and `cursor` continues from there. When pagination.has_more is true, fetch the next batch with `cursor`. Do NOT probe with a small limit to learn the dataset size — every page carries the same total_count/has_more metadata, so ask for the rows you actually need and read the metadata of that response.

How results are delivered. 'auto' (default): rows are returned inline when they fit the response budget, otherwise they are exported to a file and data.export.download_url is returned instead — so simply request the amount of data you need via `limit`. 'inline': always inline, oversized pages are truncated and continue via cursor. 'download': always a downloadable file, even for small results; the file contains exactly the window selected by `limit`/`cursor`, so size `limit` to what is wanted (the tool's max for the whole dataset). Use 'download' when the user wants a file or when you need many rows for analysis (fetch the link in your execution environment instead of paging through the conversation).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Marketplace code: US, UK, DE, JP, etc. Must be in the Magnet allowlist.

Single seed keyword, e.g. 'apple watch bands for women'.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Opaque pagination cursor from a previous response's pagination.next_cursor. Omit for the first page. Pass it back verbatim with the same query arguments (limit may change between pages); never construct or edit it. Cursors are reusable but expire 30 minutes after being issued (pagination.cursor_expires_at) and are bound to the query that issued them.

File format when results are delivered as a download (ignored for inline delivery). 'json' (default): gzip-compressed JSON, best for programmatic analysis. 'csv': UTF-8 CSV that opens directly in Excel/Sheets - prefer it when the user wants a spreadsheet/report file. Columns are the union of row fields; nested values stay JSON-encoded.

Number of rows to return, 1-10000 (default 150); out-of-range values are rejected. Up to 1000 rows come back inline; ask for more and the window is delivered as a file instead (data.export.download_url), so size `limit` to what is actually wanted. A page can still come back shorter than `limit` when the rows are unusually large (the ~1MB inline budget); pagination.truncated says so and `cursor` continues from there. When pagination.has_more is true, fetch the next batch with `cursor`. Do NOT probe with a small limit to learn the dataset size — every page carries the same total_count/has_more metadata, so ask for the rows you actually need and read the metadata of that response.

How results are delivered. 'auto' (default): rows are returned inline when they fit the response budget, otherwise they are exported to a file and data.export.download_url is returned instead — so simply request the amount of data you need via `limit`. 'inline': always inline, oversized pages are truncated and continue via cursor. 'download': always a downloadable file, even for small results; the file contains exactly the window selected by `limit`/`cursor`, so size `limit` to what is wanted (the tool's max for the whole dataset). Use 'download' when the user wants a file or when you need many rows for analysis (fetch the link in your execution environment instead of paging through the conversation).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional historical month YYYY-MM (e.g. 2025-11). Omit for current data. One month per request; must fall within the 24-month lookback ending at the current UTC month.

Required. Keyword to fetch CPC history for: phrase text (e.g. 'wireless earbuds') or a numeric Helium 10 phrase id (digit-only input is treated as a phrase id). 1-500 characters.

Required. Amazon marketplace code. Allowed values: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AU, JP, AE, BR, SA.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional window end (inclusive), format yyyy-MM-dd (UTC day), no later than today. Pass it together with start_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional window start (inclusive), format yyyy-MM-dd (UTC day). Pass it together with end_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

10-character Amazon ASIN or Amazon product URL containing an ASIN; JSON key `asin_or_url`. Required and must not be blank. If the user has not provided an ASIN or product URL, ask them to supply it—do not infer, search for, or invent an ASIN.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Opaque pagination cursor from a previous response's pagination.next_cursor. Omit for the first page. Pass it back verbatim with the same query arguments (limit may change between pages); never construct or edit it. Cursors are reusable but expire 30 minutes after being issued (pagination.cursor_expires_at) and are bound to the query that issued them.

Number of rows to return. Defaults to 100, max 40000; out-of-range values are clamped rather than rejected. request only what the user needs — up to 500 rows per page come back inline (slightly larger asks are truncated and continue via `cursor`). When pagination.has_more is true, fetch the next batch with `cursor`. Do NOT probe with limit=1 to learn the dataset size — every page carries the same total_count/has_more metadata, so ask for the rows you actually need and read the metadata of that response.

Amazon marketplace short code (e.g. `US`, `UK`, `DE`, `JP`). Optional, case-insensitive. Use one of the marketplace codes exposed in this schema.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Sort direction sent as `sort.direction`: `asc` or `desc` (case-insensitive).

Optional sort field for `keywords_new_suggestion`; must be one of the PRD sortable fields exposed in this schema.

Required search keyword phrase to query rank history for; sent as `keyword`. Must not be blank or whitespace only. The IO API performs fuzzy matching against tracked keywords on the supplied ASIN.

Required for meaningful results: 10-char ASIN or product URL (asin_or_url). Rank history is scoped to tracked keywords on this ASIN; fuzzy keyword match runs against that ASIN's tracked set. Optional for this tool — omit for account-level tracked keywords for the marketplace.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Opaque pagination cursor from a previous response's pagination.next_cursor. Omit for the first page. Pass it back verbatim with the same query arguments (limit may change between pages); never construct or edit it. Cursors are reusable but expire 30 minutes after being issued (pagination.cursor_expires_at) and are bound to the query that issued them.

Optional end of the rank-history window. ISO date in `yyyy-MM-dd` format (UTC day boundary: `start_date` is anchored at `00:00:00` and `end_date` at `23:59:59`). When omitted, invalid, the IO API falls back to the last 30 days.

Number of rows to return. Defaults to 100, max 500; out-of-range values are clamped rather than rejected. Rows are large (~2.5KB per row, up to ~20KB for keywords with Boost enabled), so request only what the user needs — up to 500 rows per page come back inline (slightly larger asks are truncated and continue via `cursor`). When pagination.has_more is true, fetch the next batch with `cursor`. Do NOT probe with limit=1 to learn the dataset size — every page carries the same total_count/has_more metadata, so ask for the rows you actually need and read the metadata of that response.

Amazon marketplace short code (e.g. `US`, `UK`, `DE`, `JP`). Optional, case-insensitive. Use one of the marketplace codes exposed in this schema.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional start of the rank-history window. ISO date in `yyyy-MM-dd` format (UTC day boundary: `start_date` is anchored at `00:00:00` and `end_date` at `23:59:59`). When omitted, invalid, the IO API falls back to the last 30 days.

Required. Keyword to fetch sales history for: phrase text (e.g. 'wireless earbuds') or a numeric Helium 10 phrase id (digit-only input is treated as a phrase id).

Required. Amazon marketplace code. Allowed values: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AU, JP, AE, BR, SA.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional window end (inclusive), format yyyy-MM-dd (UTC day), no later than today. Pass it together with start_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional window start (inclusive), format yyyy-MM-dd (UTC day). Pass it together with end_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

Required. Keyword to fetch search frequency rank history for: phrase text (e.g. 'wireless earbuds') or a numeric Helium 10 phrase id (digit-only input is treated as a phrase id).

Required. Amazon marketplace code. Allowed values: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AU, JP, AE, BR, SA.

Required. Reporting period granularity for the rank series: weekly or monthly.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional window end (inclusive), format yyyy-MM-dd (UTC day), no later than today. Pass it together with start_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional window start (inclusive), format yyyy-MM-dd (UTC day). Pass it together with end_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

Required. Keyword to fetch search volume history for: phrase text (e.g. 'wireless earbuds') or a numeric Helium 10 phrase id (digit-only input is treated as a phrase id).

Required. Amazon marketplace code. Allowed values: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AU, JP, AE, BR, SA.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional window end (inclusive), format yyyy-MM-dd (UTC day), no later than today. Pass it together with start_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional window start (inclusive), format yyyy-MM-dd (UTC day). Pass it together with end_date, both within the last 3 years, or omit both dates for the default last 30 days. Do not pass only one date.

Alias for ``main_asin`` (accepted because callers often send ``asin``). Prefer ``main_asin``; when both are supplied ``main_asin`` wins. Leave unset if you already pass ``main_asin``.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Main Amazon ASIN to analyze (10 characters). This is the seed listing whose full profile is returned along with the three chart bundles. Required — the parameter is named ``main_asin`` (not ``asin``); you may also pass it via the ``asin`` alias.

Optional marketplace short code where the ASIN is listed. Supported: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AE, BR, AU. Defaults to US when omitted. Case-insensitive; whitespace is trimmed.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Alias for ``main_asin`` (accepted because callers often send ``asin``). Prefer ``main_asin``; when both are supplied ``main_asin`` wins. Leave unset if you already pass ``main_asin``.

Optional list of up to 10 competitor ASINs (each 10 characters). When provided, ``score_data`` gains one block per competitor and ``summary_data`` reports the cross-ASIN ``average_lqs`` plus per-metric ``asins_passed`` / ``asins_failed`` lists.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Main Amazon ASIN to score (10 characters). Its own score block (``lqs`` + per-metric pass/fail) is always returned under ``score_data.<asin>``. Required — the parameter is named ``main_asin`` (not ``asin``); you may also pass it via the ``asin`` alias.

Optional marketplace short code where the ASIN is listed. Supported: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AE, BR, AU. Defaults to US when omitted. Case-insensitive; whitespace is trimmed.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Required primary range start yyyy-MM-dd. Must be on or before date_to. Dates are yyyy-MM-dd and inclusive. MT360 uses Saturday as the weekly cutoff: Mon–Fri the latest allowed end date is the Saturday of the week before last; Saturday–Sunday may use last Saturday. Each provided range must span at least 28 inclusive days.

Required primary range end yyyy-MM-dd. Must be on or after date_from. Dates are yyyy-MM-dd and inclusive. MT360 uses Saturday as the weekly cutoff: Mon–Fri the latest allowed end date is the Saturday of the week before last; Saturday–Sunday may use last Saturday. Each provided range must span at least 28 inclusive days.

Required tab. 0=CATEGORIES (top-level Amazon categories only; there is no category_id drill-down), 1=BRANDS, 2=PRODUCTS.

Required. MT360/MI market id (from list_user_markets).

Required. DSA LeaderBoardSortsV3Enum: -totalValue = top performing (highest primary value); -growth = fastest growing; -change = largest absolute increase; change = top declining (do not prefix with '-').

Optional compare range start yyyy-MM-dd. Must be sent together with compared_date_to. If both are omitted, totalValueCompared, change, and growth are null. Dates are yyyy-MM-dd and inclusive. MT360 uses Saturday as the weekly cutoff: Mon–Fri the latest allowed end date is the Saturday of the week before last; Saturday–Sunday may use last Saturday. Each provided range must span at least 28 inclusive days. When a compare range is sent (compared_date_from + compared_date_to), the two ranges must be similar length (inclusive day counts differ by at most 5).

Optional compare range end yyyy-MM-dd. Must be sent together with compared_date_from. If both are omitted, totalValueCompared, change, and growth are null. Dates are yyyy-MM-dd and inclusive. MT360 uses Saturday as the weekly cutoff: Mon–Fri the latest allowed end date is the Saturday of the week before last; Saturday–Sunday may use last Saturday. Each provided range must span at least 28 inclusive days. When a compare range is sent (compared_date_from + compared_date_to), the two ranges must be similar length (inclusive day counts differ by at most 5).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Parent vs child aggregation. Accepts boolean or integer: false/0 = child (variation) mode; true/1 = parent mode. If omitted or null, DSA defaults to child (0). Pass the market's parentDataMode from list_user_markets unless the user explicitly wants the other mode.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Required primary range start yyyy-MM-dd. Must be on or before date_to. Dates are yyyy-MM-dd and inclusive. MT360 uses Saturday as the weekly cutoff: Mon–Fri the latest allowed end date is the Saturday of the week before last; Saturday–Sunday may use last Saturday. Each provided range must span at least 28 inclusive days.

Required primary range end yyyy-MM-dd. Must be on or after date_from. Dates are yyyy-MM-dd and inclusive. MT360 uses Saturday as the weekly cutoff: Mon–Fri the latest allowed end date is the Saturday of the week before last; Saturday–Sunday may use last Saturday. Each provided range must span at least 28 inclusive days.

0 NO_GROUP, 1 PRODUCT, 2 BRAND, 3 COMPETITORS, 4 SELLER_TYPE. Do not use 0 with overall_show_type 4 (PRICE); DSA throws Wrong Params Combination.

Required. MT360/MI market id (from list_user_markets).

0 REVENUE, 1 REVENUE_MARKET_SHARE, 2 UNITS, 3 UNITS_MARKET_SHARE, 4 PRICE. PRICE (4) cannot be combined with group_by_type 0 (NO_GROUP); use 1–4. PRICE (4) is not allowed when parent_data_mode is true/1: parent aggregation has no meaningful price.

0 WEEKLY, 1 MONTHLY, 2 YEARLY.

Optional compare range start yyyy-MM-dd. Must be sent together with compared_date_to; omit both to skip the compare period. Dates are yyyy-MM-dd and inclusive. MT360 uses Saturday as the weekly cutoff: Mon–Fri the latest allowed end date is the Saturday of the week before last; Saturday–Sunday may use last Saturday. Each provided range must span at least 28 inclusive days. When a compare range is sent (compared_date_from + compared_date_to), the two ranges must be similar length (inclusive day counts differ by at most 5).

Optional compare range end yyyy-MM-dd. Must be sent together with compared_date_from; omit both to skip the compare period. Dates are yyyy-MM-dd and inclusive. MT360 uses Saturday as the weekly cutoff: Mon–Fri the latest allowed end date is the Saturday of the week before last; Saturday–Sunday may use last Saturday. Each provided range must span at least 28 inclusive days. When a compare range is sent (compared_date_from + compared_date_to), the two ranges must be similar length (inclusive day counts differ by at most 5).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

1-based page for grouped tables.

Parent vs child aggregation. Accepts boolean or integer: false/0 = child (variation) mode; true/1 = parent mode. If omitted or null, DSA defaults to child (0). Pass the market's parentDataMode from list_user_markets unless the user explicitly wants the other mode. On get_market_overall, true/1 (parent) cannot be combined with overall_show_type 4 (PRICE); parent mode has no meaningful price.

Page size for grouped tables.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Required primary range start yyyy-MM-dd. Must be on or before date_to. Dates are yyyy-MM-dd and inclusive. MT360 uses Saturday as the weekly cutoff: Mon–Fri the latest allowed end date is the Saturday of the week before last; Saturday–Sunday may use last Saturday. Each provided range must span at least 28 inclusive days.

Required primary range end yyyy-MM-dd. Must be on or after date_from. Dates are yyyy-MM-dd and inclusive. MT360 uses Saturday as the weekly cutoff: Mon–Fri the latest allowed end date is the Saturday of the week before last; Saturday–Sunday may use last Saturday. Each provided range must span at least 28 inclusive days.

Required. MT360/MI market id (from list_user_markets).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Table search string.

1-based page (default 1).

Parent vs child aggregation. Accepts boolean or integer: false/0 = child (variation) mode; true/1 = parent mode. If omitted or null, DSA defaults to child (0). Pass the market's parentDataMode from list_user_markets unless the user explicitly wants the other mode.

Rows per page (default 10, max 100).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

DSA current-data sort. Prefix '-' for descending. Default -totalRevenue when omitted. Allowed fields: price, createdAt, brand, bestMonth, bbSeller, bsr, sellersQty, FBA, fbaFee, rating, reviews, reviewsDynamic, numberOfImages, sizeTier, height, width, length, weight, variationsCount, netPrice, yearSales, salesYearOverYear, salesChange, priceChange, salesToReviews, totalSales, totalSalesActuals, totalRevenue, totalRevenueActuals. Not asin, parentAsin, or estimatedRevenue.

Required primary range start yyyy-MM-dd. Must be on or before date_to. Dates are yyyy-MM-dd and inclusive. MT360 uses Saturday as the weekly cutoff: Mon–Fri the latest allowed end date is the Saturday of the week before last; Saturday–Sunday may use last Saturday. Each provided range must span at least 28 inclusive days.

Required primary range end yyyy-MM-dd. Must be on or after date_from. Dates are yyyy-MM-dd and inclusive. MT360 uses Saturday as the weekly cutoff: Mon–Fri the latest allowed end date is the Saturday of the week before last; Saturday–Sunday may use last Saturday. Each provided range must span at least 28 inclusive days.

Required. MT360/MI market id (from list_user_markets).

Optional compare range start yyyy-MM-dd. Must be sent together with compared_date_to; omit both to skip the compare period. Dates are yyyy-MM-dd and inclusive. MT360 uses Saturday as the weekly cutoff: Mon–Fri the latest allowed end date is the Saturday of the week before last; Saturday–Sunday may use last Saturday. Each provided range must span at least 28 inclusive days. When a compare range is sent (compared_date_from + compared_date_to), the two ranges must be similar length (inclusive day counts differ by at most 5).

Optional compare range end yyyy-MM-dd. Must be sent together with compared_date_from; omit both to skip the compare period. Dates are yyyy-MM-dd and inclusive. MT360 uses Saturday as the weekly cutoff: Mon–Fri the latest allowed end date is the Saturday of the week before last; Saturday–Sunday may use last Saturday. Each provided range must span at least 28 inclusive days. When a compare range is sent (compared_date_from + compared_date_to), the two ranges must be similar length (inclusive day counts differ by at most 5).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Parent vs child aggregation. Accepts boolean or integer: false/0 = child (variation) mode; true/1 = parent mode. If omitted or null, DSA defaults to child (0). Pass the market's parentDataMode from list_user_markets unless the user explicitly wants the other mode.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Helium 10 seller identifier.

Optional ASIN. Omit for the account overview; pass an ASIN to check whether it is currently monitored.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Filter by binding validity. Omit / null (default) = return ALL sellers (active + inactive). true = only valid/active bindings. false = only invalid/inactive bindings.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Filter by binding validity. Omit / null (default) = return ALL Walmart sellers (active + inactive). true = only valid/active bindings. false = only invalid/inactive bindings. Must be a JSON boolean — 1 / 0 are INVALID_INPUT, not shorthand for true / false.

Amazon ASIN to query, 10-char alphanumeric (e.g. B01LR5S6HK).

Marketplace code: US, UK, DE, JP, etc.

Helium 10 seller identifier.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

Optional. SQL IN filter by ASIN list (array of strings, e.g. ['B001234567']). Default null (all ASINs).

Optional. Only 11 metrics support sub-component breakdown: sales, units_sold, refund, advertising_cost, amazon_fees, reimbursements, cost_of_goods, indirect_cost, gross_revenue, sessions, units_returned. The other 10 (promo, estimated_payout, gross_profit, expenses, net_profit, margin, roi, tacos, ads_acos, refunds_percentage) are derived metrics — use summary tools for those. Default null = all 11.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Optional, defaults to USD.

Timezone used to interpret calendar dates. Must be one of the Amazon marketplace report timezones (not any IANA timezone). Examples: America/Los_Angeles, Europe/Paris, Europe/London. Italy/Germany/Spain use Europe/Paris (not Europe/Rome or Europe/Berlin). Do not pass a city name, abbreviation, or UTC offset. Optional; when omitted, defaults to the queried marketplace's report timezone (America/Los_Angeles when no single marketplace is set).

Optional marketplace filter as an array of region short codes ONLY (e.g. ['US', 'CA', 'MX']); Amazon SP-API marketplace IDs (e.g. 'ATVPDKIKX0DER'), seller IDs, and domains are rejected. Allowed: AE, AU, BE, BR, CA, CN, DE, EG, ES, FR, IE, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US, ZA. Default null (all authorised marketplaces).

Optional. Page number (integer), starts from 1. Default 1.

Optional. Rows per page (integer). Min 5, max 1000. Default 50.

Row aggregation: sku | asin | parent_asin. Parent ASIN groups child ASINs; one ASIN may map to multiple SKUs.

Optional. SQL IN filter on seller account IDs (array of strings). Profits intersects with sellers from Pacvue token; no intersection → empty result. Default null (all authorised sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by seller SKU list (array of strings). Default null (all SKUs).

Optional. Metric to sort product rows by. Sort is applied to the top-level metric value per product row. If omitted, server default order applies. Allowed values: sales, units_sold, refund, advertising_cost, amazon_fees, reimbursements, cost_of_goods, indirect_cost, gross_revenue, sessions, units_returned. Cost-type metrics (refund, advertising_cost, amazon_fees, cost_of_goods, indirect_cost) are returned as negative values and are ordered by ABSOLUTE amount, not signed value: sort_order=desc lists the highest spend/cost first and asc the lowest.

Optional. Sort direction for sort_by. 'desc' = highest value first (default); 'asc' = lowest value first. Ignored if sort_by is not provided.

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

Time-bucket size for series tools: day | week | month. Use the exact parameter name granularity; do not use scope, interval, period, or bucket as input aliases. Week/month align to calendar boundaries (week starts Monday; month starts on the 1st); first/last buckets may be partial. The first bucket key may precede current_date_from (e.g. from 2026-04-01 at week granularity → first key 2026-03-30). Partial buckets include only data inside the requested date range. Series responses no longer include the scope or offset fields (removed in HP-1495); use the request granularity you sent.

Optional. SQL IN filter by ASIN list (array of strings, e.g. ['B001234567']). Default null (all ASINs).

Optional. Only 11 metrics support sub-component breakdown: sales, units_sold, refund, advertising_cost, amazon_fees, reimbursements, cost_of_goods, indirect_cost, gross_revenue, sessions, units_returned. The other 10 (promo, estimated_payout, gross_profit, expenses, net_profit, margin, roi, tacos, ads_acos, refunds_percentage) are derived metrics — use summary tools for those. Default null = all 11.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Optional, defaults to USD.

Timezone used to interpret calendar dates. Must be one of the Amazon marketplace report timezones (not any IANA timezone). Examples: America/Los_Angeles, Europe/Paris, Europe/London. Italy/Germany/Spain use Europe/Paris (not Europe/Rome or Europe/Berlin). Do not pass a city name, abbreviation, or UTC offset. Optional; when omitted, defaults to the queried marketplace's report timezone (America/Los_Angeles when no single marketplace is set).

Optional marketplace filter as an array of region short codes ONLY (e.g. ['US', 'CA', 'MX']); Amazon SP-API marketplace IDs (e.g. 'ATVPDKIKX0DER'), seller IDs, and domains are rejected. Allowed: AE, AU, BE, BR, CA, CN, DE, EG, ES, FR, IE, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US, ZA. Default null (all authorised marketplaces).

Optional. Page number (integer), starts from 1. Default 1.

Optional. Rows per page (integer). Min 5, max 1000. Default 50.

Row aggregation: sku | asin | parent_asin. Parent ASIN groups child ASINs; one ASIN may map to multiple SKUs.

Optional. SQL IN filter on seller account IDs (array of strings). Profits intersects with sellers from Pacvue token; no intersection → empty result. Default null (all authorised sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by seller SKU list (array of strings). Default null (all SKUs).

Optional. Metric to sort product rows by. Sort is applied to the top-level metric value per product row. If omitted, server default order applies. Allowed values: sales, units_sold, refund, advertising_cost, amazon_fees, reimbursements, cost_of_goods, indirect_cost, gross_revenue, sessions, units_returned. Cost-type metrics (refund, advertising_cost, amazon_fees, cost_of_goods, indirect_cost) are returned as negative values and are ordered by ABSOLUTE amount, not signed value: sort_order=desc lists the highest spend/cost first and asc the lowest.

Optional. Sort direction for sort_by. 'desc' = highest value first (default); 'asc' = lowest value first. Ignored if sort_by is not provided.

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

Optional. SQL IN filter by ASIN list (array of strings, e.g. ['B001234567']). Default null (all ASINs).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Optional, defaults to USD.

Timezone used to interpret calendar dates. Must be one of the Amazon marketplace report timezones (not any IANA timezone). Examples: America/Los_Angeles, Europe/Paris, Europe/London. Italy/Germany/Spain use Europe/Paris (not Europe/Rome or Europe/Berlin). Do not pass a city name, abbreviation, or UTC offset. Optional; when omitted, defaults to the queried marketplace's report timezone (America/Los_Angeles when no single marketplace is set).

Optional marketplace filter as an array of region short codes ONLY (e.g. ['US', 'CA', 'MX']); Amazon SP-API marketplace IDs (e.g. 'ATVPDKIKX0DER'), seller IDs, and domains are rejected. Allowed: AE, AU, BE, BR, CA, CN, DE, EG, ES, FR, IE, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US, ZA. Default null (all authorised marketplaces).

Optional. Page number (integer), starts from 1. Default 1.

Optional. Rows per page (integer). Min 5, max 1000. Default 50.

Row aggregation: sku | asin | parent_asin. Parent ASIN groups child ASINs; one ASIN may map to multiple SKUs.

Optional. SQL IN filter on seller account IDs (array of strings). Profits intersects with sellers from Pacvue token; no intersection → empty result. Default null (all authorised sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by seller SKU list (array of strings). Default null (all SKUs).

Optional. Metric to sort product rows by. Rows are ordered by the metric's value over the queried date range. If omitted, server default order applies. Allowed values: sales, units_sold, units_returned, refund, promo, advertising_cost, amazon_fees, reimbursements, estimated_payout, cost_of_goods, gross_revenue, gross_profit, expenses, indirect_cost, net_profit, margin, roi, tacos, ads_acos, refunds_percentage, sessions. Cost-type metrics (refund, promo, advertising_cost, amazon_fees, cost_of_goods, expenses, indirect_cost) are returned as negative values and are ordered by ABSOLUTE amount, not signed value: sort_order=desc lists the highest spend/cost first and asc the lowest.

Optional. Sort direction for sort_by. 'desc' = highest value first (default); 'asc' = lowest value first. Ignored if sort_by is not provided.

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

Time-bucket size for series tools: day | week | month. Use the exact parameter name granularity; do not use scope, interval, period, or bucket as input aliases. Week/month align to calendar boundaries (week starts Monday; month starts on the 1st); first/last buckets may be partial. The first bucket key may precede current_date_from (e.g. from 2026-04-01 at week granularity → first key 2026-03-30). Partial buckets include only data inside the requested date range. Series responses no longer include the scope or offset fields (removed in HP-1495); use the request granularity you sent.

Optional. SQL IN filter by ASIN list (array of strings, e.g. ['B001234567']). Default null (all ASINs).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Optional, defaults to USD.

Timezone used to interpret calendar dates. Must be one of the Amazon marketplace report timezones (not any IANA timezone). Examples: America/Los_Angeles, Europe/Paris, Europe/London. Italy/Germany/Spain use Europe/Paris (not Europe/Rome or Europe/Berlin). Do not pass a city name, abbreviation, or UTC offset. Optional; when omitted, defaults to the queried marketplace's report timezone (America/Los_Angeles when no single marketplace is set).

Optional marketplace filter as an array of region short codes ONLY (e.g. ['US', 'CA', 'MX']); Amazon SP-API marketplace IDs (e.g. 'ATVPDKIKX0DER'), seller IDs, and domains are rejected. Allowed: AE, AU, BE, BR, CA, CN, DE, EG, ES, FR, IE, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US, ZA. Default null (all authorised marketplaces).

Optional. Page number (integer), starts from 1. Default 1.

Optional. Rows per page (integer). Min 5, max 1000. Default 50.

Row aggregation: sku | asin | parent_asin. Parent ASIN groups child ASINs; one ASIN may map to multiple SKUs.

Optional. SQL IN filter on seller account IDs (array of strings). Profits intersects with sellers from Pacvue token; no intersection → empty result. Default null (all authorised sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by seller SKU list (array of strings). Default null (all SKUs).

Optional. Metric to sort product rows by. Rows are ordered by the metric's value over the queried date range. If omitted, server default order applies. Allowed values: sales, units_sold, units_returned, refund, promo, advertising_cost, amazon_fees, reimbursements, estimated_payout, cost_of_goods, gross_revenue, gross_profit, expenses, indirect_cost, net_profit, margin, roi, tacos, ads_acos, refunds_percentage, sessions. Cost-type metrics (refund, promo, advertising_cost, amazon_fees, cost_of_goods, expenses, indirect_cost) are returned as negative values and are ordered by ABSOLUTE amount, not signed value: sort_order=desc lists the highest spend/cost first and asc the lowest.

Optional. Sort direction for sort_by. 'desc' = highest value first (default); 'asc' = lowest value first. Ignored if sort_by is not provided.

Optional. SQL IN filter by ASIN list (array of strings, e.g. ['B001234567']). Default null (all ASINs).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional inclusive start day in date_tz, YYYY-MM-DD (no time). Omit BOTH current_date_from and current_date_to to default to the most recent 30 calendar days (current_date_to = today in date_tz, current_date_from = today - 29). Providing only one of the two endpoints is an error (INVALID_INPUT).

Optional inclusive end day in date_tz, YYYY-MM-DD (no time). Omit BOTH endpoints to default to the most recent 30 calendar days. Providing only one of the two endpoints is an error (INVALID_INPUT). A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

Timezone used to interpret calendar dates. Must be one of the Amazon marketplace report timezones (not any IANA timezone). Examples: America/Los_Angeles, Europe/Paris, Europe/London. Italy/Germany/Spain use Europe/Paris (not Europe/Rome or Europe/Berlin). Do not pass a city name, abbreviation, or UTC offset. Optional; when omitted, defaults to the queried marketplace's report timezone (America/Los_Angeles when no single marketplace is set).

Optional fulfillment channel filter: FBA | FBM. Omit to query both FBA and FBM inventory.

Optional time bucket size: day | week | month. Omit to default to 'day'. Week/month align to calendar boundaries (Mon start / 1st); first/last buckets may be partial.

Optional marketplace filter as an array of region short codes ONLY (e.g. ['US', 'CA', 'MX']); Amazon SP-API marketplace IDs (e.g. 'ATVPDKIKX0DER'), seller IDs, and domains are rejected. Allowed: AE, AU, BE, BR, CA, CN, DE, EG, ES, FR, IE, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US, ZA. Default null (all authorised marketplaces).

Optional. Page number (integer), starts from 1. Default 1.

Optional. Rows per page (integer). Min 5, max 1000. Default 50.

Optional. SQL IN filter on seller account IDs (array of strings). Profits intersects with sellers from Pacvue token; no intersection → empty result. Default null (all authorised sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by seller SKU list (array of strings). Default null (all SKUs).

1 to 20 Amazon ASINs to aggregate (each exactly 10 uppercase letters or digits, e.g. B07XYZ1234).

Amazon marketplace code. Must be one of: US, CA, MX, BR, DE, ES, FR, IT, UK, NL, TR, IN, AE, EG, SA, SE, PL, JP, AU, SG, BE, ZA, IE. Must be a marketplace in which the given seller_id is visible to the caller's Helium 10 account; if not, returns INTEGRATION.SELLER_NOT_ELIGIBLE. When obtained via get_brand_seller_profiles, must be one value from that row's marketplaces[].

Amazon Seller Central seller ID (pattern ^A[A-Z0-9]{9,20}$), e.g. 'A1B2C3D4E5XYZ'. May be provided directly by the user or obtained from get_brand_seller_profiles.data.sellers[].seller_id. Must belong to a seller linked to the caller's Helium 10 account and enrolled in Amazon Brand Registry; if it is not linked, not Brand-Registered, or the marketplace is not available for this seller, returns INTEGRATION.SELLER_NOT_ELIGIBLE.

Aggregation period type: 'week' (ISO week labels in time_period) or 'month' (YYYY-MM labels).

1 to 12 time periods to aggregate. When time_dimension is 'week', use ISO week labels (e.g. 2026-W18). When time_dimension is 'month', use YYYY-MM (e.g. 2026-04). Multiple periods are aggregated server-side into one row per search_query.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Opaque pagination cursor from a previous response's pagination.next_cursor. Omit for the first page. Pass it back verbatim with the same query arguments (limit may change between pages); never construct or edit it. Cursors are reusable but expire 30 minutes after being issued (pagination.cursor_expires_at) and are bound to the query that issued them.

File format when results are delivered as a download (ignored for inline delivery). 'json' (default): gzip-compressed JSON, best for programmatic analysis. 'csv': UTF-8 CSV that opens directly in Excel/Sheets - prefer it when the user wants a spreadsheet/report file. Columns are the union of row fields; nested values stay JSON-encoded.

Optional projection: return only these row fields, to save tokens when you need a few columns. Unknown names get a self-healing error listing the available fields. Overrides detail_level when both are set. Omit for the full field set.

Number of rows to return. Defaults to 50, max 10000; out-of-range values are clamped rather than rejected. Rows are large (~1KB per row), so request only what the user needs — up to 500 rows per page come back inline (slightly larger asks are truncated and continue via `cursor`); requests far above the inline budget (>1000 rows or very large payloads) are delivered as a download link (response_format=auto). When pagination.has_more is true, fetch the next batch with `cursor`. Do NOT probe with limit=1 to learn the dataset size — every page carries the same total_count/has_more metadata, so ask for the rows you actually need and read the metadata of that response.

How results are delivered. 'auto' (default): rows are returned inline when they fit the response budget, otherwise they are exported to a file and data.export.download_url is returned instead — so simply request the amount of data you need via `limit`. 'inline': always inline, oversized pages are truncated and continue via cursor. 'download': always a downloadable file, even for small results; the file contains exactly the window selected by `limit`/`cursor`, so size `limit` to what is wanted (the tool's max for the whole dataset). Use 'download' when the user wants a file or when you need many rows for analysis (fetch the link in your execution environment instead of paging through the conversation).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Marketplace code: US, UK, DE, JP, etc.

Target seller ID; can be the caller's own seller or a competitor's.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Feedback lookback window. Defaults to 90d.

Marketplace code: US, UK, DE, JP, etc.

Helium 10 seller identifier.

Optional Amazon ASIN, 10-char alphanumeric (e.g. B01LR5S6HK). Omit for all monitored products.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

The TikTok Shop product to look up collaborating influencers for, taken from a search_tiktok_products result's id.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Opaque pagination cursor from a previous response's pagination.next_cursor. Omit for the first page. Pass it back verbatim with the same query arguments (limit may change between pages); never construct or edit it. Cursors are reusable but expire 30 minutes after being issued (pagination.cursor_expires_at) and are bound to the query that issued them.

Time window in days used to calculate the metrics.

Number of rows to return, 10-100 (default 10); out-of-range values are rejected. At most 500 rows are reachable in total for one query. When pagination.has_more is true, fetch the next batch with `cursor`. Do NOT probe with a small limit to learn the dataset size — every page carries the same total_count/has_more metadata, so ask for the rows you actually need and read the metadata of that response.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Sort ascending instead of the default descending order.

Field to sort the results by.

Alias for ``main_asin`` (accepted because callers often send ``asin``). Prefer ``main_asin``; when both are supplied ``main_asin`` wins. Leave unset if you already pass ``main_asin``.

Optional list of up to 10 competitor ASINs (each 10 characters). When non-empty the response switches to multi mode: per-phrase rows gain shared-rank metrics and the aggregates move to summary_multi (Shared* variants). When omitted or empty the call runs in single mode.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Main Amazon ASIN whose keyword set anchors the response (10 characters). Required — the parameter is named ``main_asin`` (not ``asin``); you may also pass it via the ``asin`` alias.

Optional marketplace short code where the ASIN is listed. Supported: US, CA, MX, DE, ES, IT, FR, UK, IN, NL, AE, BR, AU. Defaults to US when omitted. Case-insensitive; whitespace is trimmed.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Required, non-empty. Which Level 1 parents to expand into component trees; pass only the keys you need. Allowed (Epic HP-1781 §D): cost_of_goods (Product Cost / Shipping Cost / Tariffs), indirect_cost (dynamic configured cost names), gross_revenue (Sales / Shipping / Taxes → Tax Collected + Tax Withheld), walmart_fees (Commission). Any other key → INVALID_INPUT.

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Bound to the Walmart marketplace; today only USD (US). Optional, defaults to USD.

Timezone used to interpret calendar dates. Bound to the Walmart marketplace (not any IANA timezone): today only America/Los_Angeles (US). Optional; defaults to America/Los_Angeles.

Optional. Exact IN filter by Walmart item_id list (array of strings). Narrows the account roll-up only — the response stays a single metrics map, never product rows. Default null (all items).

Optional Walmart marketplace filter as an array of country short codes. Only US is enabled today (CA / MX reserved); default null = all authorised Walmart marketplaces.

Optional. Filter on Walmart seller accounts (array of partner_id strings). Intersected with the sellers authorised on the Pacvue token; no intersection → empty metrics object. Default null (all authorised Walmart sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. Exact IN filter by seller SKU list (array of strings). Narrows the account roll-up only — the response stays a single metrics map, never product rows. Default null (all SKUs).

Required, non-empty. Which Level 1 parents to expand into component trees; pass only the keys you need. Allowed (Epic HP-1781 §D): cost_of_goods (Product Cost / Shipping Cost / Tariffs), indirect_cost (dynamic configured cost names), gross_revenue (Sales / Shipping / Taxes → Tax Collected + Tax Withheld), walmart_fees (Commission). Any other key → INVALID_INPUT.

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

Time-bucket size for series tools: day | week | month. Required. Use the exact parameter name granularity; do not use scope, interval, period, or bucket as input aliases. Week/month align to calendar boundaries (week starts Monday; month starts on the 1st); first/last buckets may be partial. The first bucket key may precede current_date_from (e.g. from 2026-07-01 at week granularity → first key 2026-06-29). Partial buckets include only data inside the requested date range. Series responses carry no scope or offset fields; use the request granularity you sent.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Bound to the Walmart marketplace; today only USD (US). Optional, defaults to USD.

Timezone used to interpret calendar dates. Bound to the Walmart marketplace (not any IANA timezone): today only America/Los_Angeles (US). Optional; defaults to America/Los_Angeles.

Optional. Exact IN filter by Walmart item_id list (array of strings). Narrows the account roll-up only — the response stays a single metrics map, never product rows. Default null (all items).

Optional Walmart marketplace filter as an array of country short codes. Only US is enabled today (CA / MX reserved); default null = all authorised Walmart marketplaces.

Optional. Filter on Walmart seller accounts (array of partner_id strings). Intersected with the sellers authorised on the Pacvue token; no intersection → empty metrics object. Default null (all authorised Walmart sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. Exact IN filter by seller SKU list (array of strings). Narrows the account roll-up only — the response stays a single metrics map, never product rows. Default null (all SKUs).

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Bound to the Walmart marketplace; today only USD (US). Optional, defaults to USD.

Timezone used to interpret calendar dates. Bound to the Walmart marketplace (not any IANA timezone): today only America/Los_Angeles (US). Optional; defaults to America/Los_Angeles.

Optional. Exact IN filter by Walmart item_id list (array of strings). Narrows the account roll-up only — the response stays a single metrics map, never product rows. Default null (all items).

Optional Walmart marketplace filter as an array of country short codes. Only US is enabled today (CA / MX reserved); default null = all authorised Walmart marketplaces.

Optional. Filter on Walmart seller accounts (array of partner_id strings). Intersected with the sellers authorised on the Pacvue token; no intersection → empty metrics object. Default null (all authorised Walmart sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. Exact IN filter by seller SKU list (array of strings). Narrows the account roll-up only — the response stays a single metrics map, never product rows. Default null (all SKUs).

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

Time-bucket size for series tools: day | week | month. Required. Use the exact parameter name granularity; do not use scope, interval, period, or bucket as input aliases. Week/month align to calendar boundaries (week starts Monday; month starts on the 1st); first/last buckets may be partial. The first bucket key may precede current_date_from (e.g. from 2026-07-01 at week granularity → first key 2026-06-29). Partial buckets include only data inside the requested date range. Series responses carry no scope or offset fields; use the request granularity you sent.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Bound to the Walmart marketplace; today only USD (US). Optional, defaults to USD.

Timezone used to interpret calendar dates. Bound to the Walmart marketplace (not any IANA timezone): today only America/Los_Angeles (US). Optional; defaults to America/Los_Angeles.

Optional. Exact IN filter by Walmart item_id list (array of strings). Narrows the account roll-up only — the response stays a single metrics map, never product rows. Default null (all items).

Optional Walmart marketplace filter as an array of country short codes. Only US is enabled today (CA / MX reserved); default null = all authorised Walmart marketplaces.

Optional. Filter on Walmart seller accounts (array of partner_id strings). Intersected with the sellers authorised on the Pacvue token; no intersection → empty metrics object. Default null (all authorised Walmart sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. Exact IN filter by seller SKU list (array of strings). Narrows the account roll-up only — the response stays a single metrics map, never product rows. Default null (all SKUs).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional. Filter by the product's listing attribute: WFS (Walmart Fulfillment Services) | WFSE (WFS Eligible) | SF (Seller Fulfilled). Multi-select ORs within the array; omit / null = all three. NOTE: this matches the product's fulfillment attribute — filtered rows may STILL show non-zero quantities on the other channel (e.g. fulfillment_type=['WFS'] can return rows with sf_available > 0), because one Walmart SKU can hold WFS and SF stock at the same time.

Optional. Exact IN filter by Walmart item_id list (array of strings). Restricts which product rows are returned; the response is still one row per product. Default null (all items).

Optional Walmart marketplace filter as an array of country short codes. Only US is enabled today (CA / MX reserved); default null = all authorised Walmart marketplaces.

Optional. 1-based page number. Default 1.

Optional. Rows per page, 5..1000. Default 50.

Optional. Filter on Walmart seller accounts (array of partner_id strings). Intersected with the sellers authorised on the Pacvue token; no intersection → empty metrics object. Default null (all authorised Walmart sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional inclusive maximum bound on sf_available (non-negative integer). Set only sf_available_min for >= min, only sf_available_max for <= max, or both for a closed range; sf_available_min must be <= sf_available_max. Range filters on different metrics are ANDed.

Optional inclusive minimum bound on sf_available (non-negative integer). Set only sf_available_min for >= min, only sf_available_max for <= max, or both for a closed range; sf_available_min must be <= sf_available_max. Range filters on different metrics are ANDed.

Optional inclusive maximum bound on sf_on_hand (non-negative integer). Set only sf_on_hand_min for >= min, only sf_on_hand_max for <= max, or both for a closed range; sf_on_hand_min must be <= sf_on_hand_max. Range filters on different metrics are ANDed.

Optional inclusive minimum bound on sf_on_hand (non-negative integer). Set only sf_on_hand_min for >= min, only sf_on_hand_max for <= max, or both for a closed range; sf_on_hand_min must be <= sf_on_hand_max. Range filters on different metrics are ANDed.

Optional. Exact IN filter by seller SKU list (array of strings). Restricts which product rows are returned; the response is still one row per product. Default null (all SKUs).

Optional. Sort rows by one of the seven quantity metrics. Omit = server default order. Combine with a small page_size to get top/bottom N SKUs in one call (e.g. sort_by=wfs_available + sort_order=asc surfaces lowest-stock SKUs first for restock triage).

Optional. Sort direction for sort_by. 'desc' = highest value first (default); 'asc' = lowest value first. Ignored if sort_by is not provided.

Optional inclusive maximum bound on wfs_available (non-negative integer). Set only wfs_available_min for >= min, only wfs_available_max for <= max, or both for a closed range; wfs_available_min must be <= wfs_available_max. Range filters on different metrics are ANDed.

Optional inclusive minimum bound on wfs_available (non-negative integer). Set only wfs_available_min for >= min, only wfs_available_max for <= max, or both for a closed range; wfs_available_min must be <= wfs_available_max. Range filters on different metrics are ANDed.

Optional inclusive maximum bound on wfs_inbound (non-negative integer). Set only wfs_inbound_min for >= min, only wfs_inbound_max for <= max, or both for a closed range; wfs_inbound_min must be <= wfs_inbound_max. Range filters on different metrics are ANDed.

Optional inclusive minimum bound on wfs_inbound (non-negative integer). Set only wfs_inbound_min for >= min, only wfs_inbound_max for <= max, or both for a closed range; wfs_inbound_min must be <= wfs_inbound_max. Range filters on different metrics are ANDed.

Optional inclusive maximum bound on wfs_on_hand (non-negative integer). Set only wfs_on_hand_min for >= min, only wfs_on_hand_max for <= max, or both for a closed range; wfs_on_hand_min must be <= wfs_on_hand_max. Range filters on different metrics are ANDed.

Optional inclusive minimum bound on wfs_on_hand (non-negative integer). Set only wfs_on_hand_min for >= min, only wfs_on_hand_max for <= max, or both for a closed range; wfs_on_hand_min must be <= wfs_on_hand_max. Range filters on different metrics are ANDed.

Optional inclusive maximum bound on wfs_reserved (non-negative integer). Set only wfs_reserved_min for >= min, only wfs_reserved_max for <= max, or both for a closed range; wfs_reserved_min must be <= wfs_reserved_max. Range filters on different metrics are ANDed.

Optional inclusive minimum bound on wfs_reserved (non-negative integer). Set only wfs_reserved_min for >= min, only wfs_reserved_max for <= max, or both for a closed range; wfs_reserved_min must be <= wfs_reserved_max. Range filters on different metrics are ANDed.

Optional inclusive maximum bound on wfs_unavailable (non-negative integer). Set only wfs_unavailable_min for >= min, only wfs_unavailable_max for <= max, or both for a closed range; wfs_unavailable_min must be <= wfs_unavailable_max. Range filters on different metrics are ANDed.

Optional inclusive minimum bound on wfs_unavailable (non-negative integer). Set only wfs_unavailable_min for >= min, only wfs_unavailable_max for <= max, or both for a closed range; wfs_unavailable_min must be <= wfs_unavailable_max. Range filters on different metrics are ANDed.

Required, non-empty. Which Level 1 parents to expand into component trees; pass only the keys you need. Allowed (Epic HP-1781 §D): cost_of_goods (Product Cost / Shipping Cost / Tariffs), indirect_cost (dynamic configured cost names), gross_revenue (Sales / Shipping / Taxes → Tax Collected + Tax Withheld), walmart_fees (Commission). Any other key → INVALID_INPUT.

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Bound to the Walmart marketplace; today only USD (US). Optional, defaults to USD.

Timezone used to interpret calendar dates. Bound to the Walmart marketplace (not any IANA timezone): today only America/Los_Angeles (US). Optional; defaults to America/Los_Angeles.

Optional. Exact IN filter by Walmart item_id list (array of strings). Restricts which product rows are returned; the response is still one row per product. Default null (all items).

Optional Walmart marketplace filter as an array of country short codes. Only US is enabled today (CA / MX reserved); default null = all authorised Walmart marketplaces.

Optional. 1-based page number. Default 1.

Optional. Rows per page, 1..100. Default 50.

Row aggregation for product tools: sku (default) | item_id. One item_id may map to multiple SKUs. Walmart has no asin/parent_asin levels.

Optional. Filter on Walmart seller accounts (array of partner_id strings). Intersected with the sellers authorised on the Pacvue token; no intersection → empty metrics object. Default null (all authorised Walmart sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. Exact IN filter by seller SKU list (array of strings). Restricts which product rows are returned; the response is still one row per product. Default null (all SKUs).

Optional. Sort rows by any Walmart Level 1 metric key (Epic HP-1781 §C). Cost-type metrics (refund, advertising_cost, cost_of_goods, indirect_cost, walmart_fees, expenses) are returned as negative values and are ordered by ABSOLUTE amount, not signed value: sort_order=desc lists the highest spend/cost first and asc the lowest — same rule as the Amazon P&L tools. net_profit, margin and roi can also be negative but are NOT cost metrics: they keep signed ordering, so desc lists the most profitable first and losses last. Omit = backend default order.

Optional. Sort direction for sort_by. 'desc' = highest value first (default); 'asc' = lowest value first. Ignored if sort_by is not provided.

Required, non-empty. Which Level 1 parents to expand into component trees; pass only the keys you need. Allowed (Epic HP-1781 §D): cost_of_goods (Product Cost / Shipping Cost / Tariffs), indirect_cost (dynamic configured cost names), gross_revenue (Sales / Shipping / Taxes → Tax Collected + Tax Withheld), walmart_fees (Commission). Any other key → INVALID_INPUT.

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

Time-bucket size for series tools: day | week | month. Required. Use the exact parameter name granularity; do not use scope, interval, period, or bucket as input aliases. Week/month align to calendar boundaries (week starts Monday; month starts on the 1st); first/last buckets may be partial. The first bucket key may precede current_date_from (e.g. from 2026-07-01 at week granularity → first key 2026-06-29). Partial buckets include only data inside the requested date range. Series responses carry no scope or offset fields; use the request granularity you sent.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Bound to the Walmart marketplace; today only USD (US). Optional, defaults to USD.

Timezone used to interpret calendar dates. Bound to the Walmart marketplace (not any IANA timezone): today only America/Los_Angeles (US). Optional; defaults to America/Los_Angeles.

Optional. Exact IN filter by Walmart item_id list (array of strings). Restricts which product rows are returned; the response is still one row per product. Default null (all items).

Optional Walmart marketplace filter as an array of country short codes. Only US is enabled today (CA / MX reserved); default null = all authorised Walmart marketplaces.

Optional. 1-based page number. Default 1.

Optional. Rows per page, 1..100. Default 50.

Row aggregation for product tools: sku (default) | item_id. One item_id may map to multiple SKUs. Walmart has no asin/parent_asin levels.

Optional. Filter on Walmart seller accounts (array of partner_id strings). Intersected with the sellers authorised on the Pacvue token; no intersection → empty metrics object. Default null (all authorised Walmart sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. Exact IN filter by seller SKU list (array of strings). Restricts which product rows are returned; the response is still one row per product. Default null (all SKUs).

Optional. Sort rows by any Walmart Level 1 metric key (Epic HP-1781 §C). Cost-type metrics (refund, advertising_cost, cost_of_goods, indirect_cost, walmart_fees, expenses) are returned as negative values and are ordered by ABSOLUTE amount, not signed value: sort_order=desc lists the highest spend/cost first and asc the lowest — same rule as the Amazon P&L tools. net_profit, margin and roi can also be negative but are NOT cost metrics: they keep signed ordering, so desc lists the most profitable first and losses last. Omit = backend default order.

Optional. Sort direction for sort_by. 'desc' = highest value first (default); 'asc' = lowest value first. Ignored if sort_by is not provided.

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Bound to the Walmart marketplace; today only USD (US). Optional, defaults to USD.

Timezone used to interpret calendar dates. Bound to the Walmart marketplace (not any IANA timezone): today only America/Los_Angeles (US). Optional; defaults to America/Los_Angeles.

Optional. Exact IN filter by Walmart item_id list (array of strings). Restricts which product rows are returned; the response is still one row per product. Default null (all items).

Optional Walmart marketplace filter as an array of country short codes. Only US is enabled today (CA / MX reserved); default null = all authorised Walmart marketplaces.

Optional. 1-based page number. Default 1.

Optional. Rows per page, 1..100. Default 50.

Row aggregation for product tools: sku (default) | item_id. One item_id may map to multiple SKUs. Walmart has no asin/parent_asin levels.

Optional. Filter on Walmart seller accounts (array of partner_id strings). Intersected with the sellers authorised on the Pacvue token; no intersection → empty metrics object. Default null (all authorised Walmart sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. Exact IN filter by seller SKU list (array of strings). Restricts which product rows are returned; the response is still one row per product. Default null (all SKUs).

Optional. Sort rows by any Walmart Level 1 metric key (Epic HP-1781 §C). Cost-type metrics (refund, advertising_cost, cost_of_goods, indirect_cost, walmart_fees, expenses) are returned as negative values and are ordered by ABSOLUTE amount, not signed value: sort_order=desc lists the highest spend/cost first and asc the lowest — same rule as the Amazon P&L tools. net_profit, margin and roi can also be negative but are NOT cost metrics: they keep signed ordering, so desc lists the most profitable first and losses last. Omit = backend default order.

Optional. Sort direction for sort_by. 'desc' = highest value first (default); 'asc' = lowest value first. Ignored if sort_by is not provided.

Inclusive start calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_from; do not use start_date, date_from, or from_date. Must be <= current_date_to.

Inclusive end calendar date in date_tz, YYYY-MM-DD only (no time). Required. Use the exact parameter name current_date_to; do not use end_date, date_to, or to_date. Must be >= current_date_from. A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

Time-bucket size for series tools: day | week | month. Required. Use the exact parameter name granularity; do not use scope, interval, period, or bucket as input aliases. Week/month align to calendar boundaries (week starts Monday; month starts on the 1st); first/last buckets may be partial. The first bucket key may precede current_date_from (e.g. from 2026-07-01 at week granularity → first key 2026-06-29). Partial buckets include only data inside the requested date range. Series responses carry no scope or offset fields; use the request granularity you sent.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

ISO 4217 code for amounts. Bound to the Walmart marketplace; today only USD (US). Optional, defaults to USD.

Timezone used to interpret calendar dates. Bound to the Walmart marketplace (not any IANA timezone): today only America/Los_Angeles (US). Optional; defaults to America/Los_Angeles.

Optional. Exact IN filter by Walmart item_id list (array of strings). Restricts which product rows are returned; the response is still one row per product. Default null (all items).

Optional Walmart marketplace filter as an array of country short codes. Only US is enabled today (CA / MX reserved); default null = all authorised Walmart marketplaces.

Optional. 1-based page number. Default 1.

Optional. Rows per page, 1..100. Default 50.

Row aggregation for product tools: sku (default) | item_id. One item_id may map to multiple SKUs. Walmart has no asin/parent_asin levels.

Optional. Filter on Walmart seller accounts (array of partner_id strings). Intersected with the sellers authorised on the Pacvue token; no intersection → empty metrics object. Default null (all authorised Walmart sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. Exact IN filter by seller SKU list (array of strings). Restricts which product rows are returned; the response is still one row per product. Default null (all SKUs).

Optional. Sort rows by any Walmart Level 1 metric key (Epic HP-1781 §C). Cost-type metrics (refund, advertising_cost, cost_of_goods, indirect_cost, walmart_fees, expenses) are returned as negative values and are ordered by ABSOLUTE amount, not signed value: sort_order=desc lists the highest spend/cost first and asc the lowest — same rule as the Amazon P&L tools. net_profit, margin and roi can also be negative but are NOT cost metrics: they keep signed ordering, so desc lists the most profitable first and losses last. Omit = backend default order.

Optional. Sort direction for sort_by. 'desc' = highest value first (default); 'asc' = lowest value first. Ignored if sort_by is not provided.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional inclusive start day in date_tz, YYYY-MM-DD (no time). Omit BOTH endpoints to default to the most recent 30 calendar days. Providing only one of the two endpoints is an error (INVALID_INPUT).

Optional inclusive end day in date_tz, YYYY-MM-DD (no time). Omit BOTH endpoints to default to the most recent 30 calendar days. Providing only one of the two endpoints is an error (INVALID_INPUT). A value after today in date_tz is silently clamped to today in date_tz (the future has no data).

Timezone used to interpret calendar dates. Bound to the Walmart marketplace (not any IANA timezone): today only America/Los_Angeles (US). Optional; defaults to America/Los_Angeles.

Optional. Filter by the product's listing attribute: WFS (Walmart Fulfillment Services) | WFSE (WFS Eligible) | SF (Seller Fulfilled). Multi-select ORs within the array; omit / null = all three. NOTE: this matches the product's fulfillment attribute — filtered rows may STILL show non-zero quantities on the other channel (e.g. fulfillment_type=['WFS'] can return rows with sf_available > 0), because one Walmart SKU can hold WFS and SF stock at the same time.

Optional time-bucket size: day | week | month. Omit to default to 'day'. Week/month align to calendar boundaries (week starts Monday; month on the 1st); first/last buckets may be partial vs the requested range. Bucket keys match the Walmart P&L series tools for the same inputs.

Optional. Exact IN filter by Walmart item_id list (array of strings). Restricts which product rows are returned; the response is still one row per product. Default null (all items).

Optional Walmart marketplace filter as an array of country short codes. Only US is enabled today (CA / MX reserved); default null = all authorised Walmart marketplaces.

Optional. 1-based page number. Default 1.

Optional. Rows per page, 5..1000. Default 50.

Optional. Filter on Walmart seller accounts (array of partner_id strings). Intersected with the sellers authorised on the Pacvue token; no intersection → empty metrics object. Default null (all authorised Walmart sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. Exact IN filter by seller SKU list (array of strings). Restricts which product rows are returned; the response is still one row per product. Default null (all SKUs).

Platform key from get_ads_rule_platforms. Required. v1 supports amazon only (walmart planned). Case-sensitive.

Rule type from get_ads_rule_platforms. Required -- ask the user which rule type to list. v1 supports BID only. Case-sensitive.

Which collection to list: 'live' = committed automation rules; 'draft' = the draft box (saved, not yet committed).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Max items per page. Default 50, max 500.

Case-insensitive substring filter on the rule name.

Zero-based paging offset. Default 0.

Target profile_id (from get_ads_query_materials). Optional for source=draft; for source=live it scopes the tenant -- omit it only if you want the server to auto-resolve your single accessible profile.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by ASIN list (array of strings, e.g. ['B001234567']). Default null (all ASINs).

Filter by COGS configuration type: by_period | persistent. Omit = both types.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

yyyy-MM-dd. Keep rows whose ends_at is <= this date (inclusive). Note: open-ended by_period rows (ends_at=null) and persistent rows never match any ends_at_* filter. Default: null (no bound).

yyyy-MM-dd. Keep rows whose ends_at is >= this date (inclusive). Note: open-ended by_period rows (ends_at=null) and persistent rows never match any ends_at_* filter. Default: null (no bound).

Filter on the SKU fulfillment channel: FBA | FBM. Omit = both.

Optional marketplace filter as an array of upper-case region short codes (e.g. ['US', 'CA', 'MX']). Case-sensitive — lower-case codes are rejected. Allowed: AE, AU, BE, BR, CA, CN, DE, EG, ES, FR, IE, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US, ZA. Default null (all authorised marketplaces).

Optional. Page number (integer), starts from 1. Default 1.

Optional. Rows per page (integer). Min 5, max 1000. Default 50.

Inclusive upper bound on product_cost (>= 0). Compared as stored in each row's currency — no FX conversion across marketplaces. Default: null (no bound).

Inclusive lower bound on product_cost (>= 0). Compared as stored in each row's currency — no FX conversion across marketplaces. Default: null (no bound).

Optional. SQL IN filter on seller account IDs (array of strings). Profits intersects with sellers from Pacvue token; no intersection → empty result. Default null (all authorised sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Inclusive upper bound on shipping_cost (>= 0). Compared as stored in each row's currency — no FX conversion across marketplaces. Default: null (no bound).

Inclusive lower bound on shipping_cost (>= 0). Compared as stored in each row's currency — no FX conversion across marketplaces. Default: null (no bound).

Amazon listing status on the SKU (active | inactive) — a SKU attribute, not a COGS attribute. Omit = all statuses.

Optional. SQL IN filter by seller SKU list (array of strings). Default null (all SKUs).

Field to sort rows by. Allowlist: product_cost | shipping_cost | tariffs | total_cogs | starts_at | ends_at. Omit = server default stable order. When sorting on starts_at/ends_at, rows with null in that field are ordered after non-null rows.

Sort direction: asc | desc (default desc — highest/latest first). Ignored when sort_by is omitted.

yyyy-MM-dd. Keep rows whose starts_at is <= this date (inclusive, day precision). Default: null (no bound).

yyyy-MM-dd. Keep rows whose starts_at is >= this date (inclusive, day precision). Default: null (no bound).

Inclusive upper bound on tariffs (>= 0). Compared as stored in each row's currency — no FX conversion across marketplaces. Default: null (no bound).

Inclusive lower bound on tariffs (>= 0). Compared as stored in each row's currency — no FX conversion across marketplaces. Default: null (no bound).

Inclusive upper bound on total_cogs (>= 0). Compared as stored in each row's currency — no FX conversion across marketplaces. Default: null (no bound).

Inclusive lower bound on total_cogs (>= 0). Compared as stored in each row's currency — no FX conversion across marketplaces. Default: null (no bound).

Inclusive upper bound on the setting amount (> 0). Compared as stored (no FX). Default: null.

Inclusive lower bound on the setting amount (> 0). Compared as stored (no FX). Default: null.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Filter by inferred level: h10_account | seller | marketplace | product. Omit = all.

Case-sensitive substring match (LIKE) on the setting name. Omit = no name filter.

yyyy-MM-dd. Keep settings whose created_at calendar day is <= this (inclusive).

yyyy-MM-dd. Keep settings whose created_at calendar day is >= this (inclusive).

yyyy-MM-dd. Keep settings whose ends_at is <= this (inclusive). Open-ended settings (ends_at null) never match an ends_at_* filter.

yyyy-MM-dd. Keep settings whose ends_at is >= this (inclusive). Open-ended settings (ends_at null) never match an ends_at_* filter.

Filter by frequency: one_time | daily | weekly | monthly | quarterly | half_yearly | annually. Omit = all.

Optional marketplace filter as an array of upper-case region short codes (e.g. ['US', 'CA', 'MX']). Case-sensitive — lower-case codes are rejected. Allowed: AE, AU, BE, BR, CA, CN, DE, EG, ES, FR, IE, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US, ZA. Default null (all authorised marketplaces).

Optional. Page number (integer), starts from 1. Default 1.

Optional. Rows per page (integer). Min 5, max 1000. Default 50.

Optional. SQL IN filter on seller account IDs (array of strings). Profits intersects with sellers from Pacvue token; no intersection → empty result. Default null (all authorised sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by seller SKU list (array of strings). Default null (all SKUs).

Field to sort by: amount | created_at | starts_at | ends_at. Omit = server default.

Sort direction: asc | desc (default desc). Ignored when sort_by is omitted.

yyyy-MM-dd. Keep settings whose starts_at is <= this (inclusive).

yyyy-MM-dd. Keep settings whose starts_at is >= this (inclusive).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Filter by inferred level: h10_account | seller | marketplace | product. Omit = all.

Case-sensitive substring match (LIKE) on the setting name. Omit = no name filter.

Filter by setting frequency: one_time | daily | weekly | monthly | quarterly | half_yearly | annually. Omit = all.

Optional marketplace filter as an array of upper-case region short codes (e.g. ['US', 'CA', 'MX']). Case-sensitive — lower-case codes are rejected. Allowed: AE, AU, BE, BR, CA, CN, DE, EG, ES, FR, IE, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US, ZA. Default null (all authorised marketplaces).

Inclusive upper bound on a row's occurrence_amount (> 0). Compared as stored (no FX). Default: null (no bound).

Inclusive lower bound on a row's occurrence_amount (> 0). Compared as stored (no FX). Default: null (no bound).

yyyy-MM-dd. Keep charges whose occurrence_date is <= this day (inclusive).

yyyy-MM-dd. Keep charges whose occurrence_date is >= this day (inclusive).

Optional. Page number (integer), starts from 1. Default 1.

Optional. Rows per page (integer). Min 5, max 1000. Default 50.

Optional. SQL IN filter on seller account IDs (array of strings). Profits intersects with sellers from Pacvue token; no intersection → empty result. Default null (all authorised sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by seller SKU list (array of strings). Default null (all SKUs).

Field to sort rows by: occurrence_date | occurrence_amount. Omit = server default.

Sort direction: asc | desc (default desc). Ignored when sort_by is omitted.

Optional. SQL IN filter by ASIN list (array of strings, e.g. ['B001234567']). Default null (all ASINs).

Optional. Case-insensitive substring match (SQL LIKE '%keyword%') against the brand name. Omit or null = no brand filter (rows with brand: null may still be returned).

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Optional fulfillment channel filter: FBA | FBM. Omit to query both FBA and FBM inventory.

Optional marketplace filter as an array of region short codes ONLY (e.g. ['US', 'CA', 'MX']); Amazon SP-API marketplace IDs (e.g. 'ATVPDKIKX0DER'), seller IDs, and domains are rejected. Allowed: AE, AU, BE, BR, CA, CN, DE, EG, ES, FR, IE, IN, IT, JP, MX, NL, PL, SA, SE, SG, TR, UK, US, ZA. Default null (all authorised marketplaces).

Optional. Page number (integer), starts from 1. Default 1.

Optional. Rows per page (integer). Min 5, max 1000. Default 50.

Optional. Case-insensitive substring match (SQL LIKE '%keyword%') against the product listing title. Omit or null = no title filter (rows with any title, including empty, may be returned).

Optional. SQL IN filter on seller account IDs (array of strings). Profits intersects with sellers from Pacvue token; no intersection → empty result. Default null (all authorised sellers).

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Optional. SQL IN filter by seller SKU list (array of strings). Default null (all SKUs).

10-character Amazon ASIN or Amazon product URL containing an ASIN; JSON key `asin_or_url`. Required and must not be blank. If the user has not provided an ASIN or product URL, ask them to supply it—do not infer, search for, or invent an ASIN.

Optional filter: when `true`, only boosted tracked keywords. When omitted or `false`, no boost-based filter is applied.

In one short sentence, explain why you are calling this tool and what you are trying to accomplish for the user. Used for analytics only; it does not affect the tool's behavior.

Opaque pagination cursor from a previous response's pagination.next_cursor. Omit for the first page. Pass it back verbatim with the same query arguments (limit may change between pages); never construct or edit it. Cursors are reusable but expire 30 minutes after being issued (pagination.cursor_expires_at) and are bound to the query that issued them.

Number of rows to return. Defaults to 100, max 500; out-of-range values are clamped rather than rejected. request only what the user needs — up to 500 rows per page come back inline (slightly larger asks are truncated and continue via `cursor`). When pagination.has_more is true, fetch the next batch with `cursor`. Do NOT probe with limit=1 to learn the dataset size — every page carries the same total_count/has_more metadata, so ask for the rows you actually need and read the metadata of that response.

Amazon marketplace short code (e.g. `US`, `UK`, `DE`, `JP`). Optional, case-insensitive. Use one of the marketplace codes exposed in this schema.

REQUIRED on every call after your first. This MCP server associates all of your tool calls for a given task using session_id; a call that arrives without it cannot be associated with your earlier work and is treated as the start of an unrelated task, disconnected from the original goal. Omit it on your first call only — the server will issue one in a trailing [MCP INSTRUCTIONS] text block of the result — then echo that exact value on every later call. Never invent a value, and do not issue parallel tool calls until the server has issued your session_id. One session_id covers the whole goal from start to finish: when you spawn subagents you MUST give them this same session_id, or their work is severed from yours. session_id is internal plumbing: never show, quote or mention it to the user. Without session_id, this server does not function as intended.

Sort direction sent as `sort.direction`: `asc` or `desc` (case-insensitive).

Optional sort field for `competitor_keywords_tracking`; must be one of the PRD sortable fields exposed in this schema.

Optional filter: Keyword Tracker tag titles (account scope), OR logic. IO resolves titles to ids; blank entries ignored. Valid tag with no matches → empty rows; unknown/invalid tag title → no data