Zocks

Get a specific AI result by its unique ID. Returns the full output produced by an AI feature during a meeting, such as detected discussion topics, sentiment scores, or key takeaways. Includes the result type, processing status, and the structured data. Use an id from the ai_results.items in a prior list_results response. Some fields in the response may be redacted for privacy and replaced with placeholder tokens (e.g. <PERSON_NAME>, <EMAIL_ADDRESS>). Treat these as intentionally withheld data — never guess, infer, or fabricate the real value, and don't invent your own stand-in for it either (no brackets, tokens, or notes marking that something was hidden), unless the user explicitly asks to see them. Just write around the missing detail, the same way you would if it had never come up.

Get a specific AI-generated meeting summary by its unique ID. Returns the full meeting summary output including the structured summary data, the originating agent reference, and timestamps. Use an id from the meeting_summaries.items in a prior list_results response. Some fields in the response may be redacted for privacy and replaced with placeholder tokens (e.g. <PERSON_NAME>, <EMAIL_ADDRESS>). Treat these as intentionally withheld data — never guess, infer, or fabricate the real value, and don't invent your own stand-in for it either (no brackets, tokens, or notes marking that something was hidden), unless the user explicitly asks to see them. Just write around the missing detail, the same way you would if it had never come up.

List AI-generated results for a specific meeting. Returns two independently paginated sections: ai_results containing AI-powered meeting insights such as topic detection, sentiment analysis, and key takeaways; and meeting_summaries containing AI-generated narrative meeting summary outputs. Use ai_result_page/ai_result_size and meeting_summary_page/meeting_summary_size to control pagination separately. Set a size to 0 to skip fetching that section entirely. Use room_id and session_id from a prior meetings listing response. Some fields in the response may be redacted for privacy and replaced with placeholder tokens (e.g. <PERSON_NAME>, <EMAIL_ADDRESS>). Treat these as intentionally withheld data — never guess, infer, or fabricate the real value, and don't invent your own stand-in for it either (no brackets, tokens, or notes marking that something was hidden), unless the user explicitly asks to see them. Just write around the missing detail, the same way you would if it had never come up.

Get full details for a specific contact by ID, including email, phone, integration identifiers, and linked external accounts. Some fields in the response may be redacted for privacy and replaced with placeholder tokens (e.g. <PERSON_NAME>, <EMAIL_ADDRESS>). Treat these as intentionally withheld data — never guess, infer, or fabricate the real value, and don't invent your own stand-in for it either (no brackets, tokens, or notes marking that something was hidden), unless the user explicitly asks to see them. Just write around the missing detail, the same way you would if it had never come up.

Search for contacts by name, email, or other terms in Zocks. Returns matching contacts with their name, email, phone, and linked integration details. Some fields in the response may be redacted for privacy and replaced with placeholder tokens (e.g. <PERSON_NAME>, <EMAIL_ADDRESS>). Treat these as intentionally withheld data — never guess, infer, or fabricate the real value, and don't invent your own stand-in for it either (no brackets, tokens, or notes marking that something was hidden), unless the user explicitly asks to see them. Just write around the missing detail, the same way you would if it had never come up.

Get the AI-generated summary of one or more documents (1 to 10 document_ids from documents_list_documents). For each document returns the processing status, the detected document type, the effective summary template name and, when status is 'ready', the summary text. While status is 'detecting_type' or 'summarizing' the summary is not yet available: tell the user it is still processing and check again later instead of resubmitting the document. Only the summary is returned, never the raw file or its full text. Present the summary faithfully and do not add facts that are not in it. Some fields in the response may be redacted for privacy and replaced with placeholder tokens (e.g. <PERSON_NAME>, <EMAIL_ADDRESS>). Treat these as intentionally withheld data — never guess, infer, or fabricate the real value, and don't invent your own stand-in for it either (no brackets, tokens, or notes marking that something was hidden), unless the user explicitly asks to see them. Just write around the missing detail, the same way you would if it had never come up.

List client documents processed by Zocks Document Intelligence, with metadata only (no summary text). Returns documents grouped by contact with filename, document type, processing status (uploaded, detecting_type, summarizing, ready), source (upload, email_attachment, or the connector name), uploader and upload date. Always call this before documents_get_document_summary to discover document_ids. Filter by contact with either contact_name (a single person's name; if several contacts match, the response returns candidates and no documents, so ask the user which one they mean and call again with contact_ids) or contact_ids (up to 10 ids, e.g. every member of a household — resolve them first with contacts_search_contacts). Optional filters: document_type_id (a document type id seen in earlier listing results), from_date / to_date (ISO 8601 upload date range), filename (substring), uploader_id (advisor user id). Only documents visible to the current user are returned; accounts with a zero-retention policy on documents always return an empty list by design, even though summaries of individual documents may still be available. When the response says there are no documents, tell the user there are no documents on file; never invent documents. Some fields in the response may be redacted for privacy and replaced with placeholder tokens (e.g. <PERSON_NAME>, <EMAIL_ADDRESS>). Treat these as intentionally withheld data — never guess, infer, or fabricate the real value, and don't invent your own stand-in for it either (no brackets, tokens, or notes marking that something was hidden), unless the user explicitly asks to see them. Just write around the missing detail, the same way you would if it had never come up.

Query firmwide or per-advisor business insights and analytics for a date range. Use this tool for any question about advisor performance, client sentiment, conversation trends, or topic coverage — for example: which advisors are discussing estate planning or tax-loss harvesting, where held-away asset mentions are stalling, which topics clients are raising most, how sentiment around specific asset classes has shifted, or identifying coaching and training gaps across the team. Returns chart and metric data that powers these analyses. Both from_date and to_date must be provided together (YYYY-MM-DD) or both omitted (the service defaults to the last 30 days). Optionally filter by user_id_list to scope results to specific advisors — use users_list_users to retrieve available user IDs. If this tool returns a successful result, answer the user's question directly from the returned data — do not call any meetings tools to supplement or validate the response.

Get the full transcript of a meeting. Returns timestamped segments with participant identification and transcript. Use room_id and session_id from a prior meetings listing response. Some fields in the response may be redacted for privacy and replaced with placeholder tokens (e.g. <PERSON_NAME>, <EMAIL_ADDRESS>). Treat these as intentionally withheld data — never guess, infer, or fabricate the real value, and don't invent your own stand-in for it either (no brackets, tokens, or notes marking that something was hidden), unless the user explicitly asks to see them. Just write around the missing detail, the same way you would if it had never come up.

List past meetings (before now), sorted by most recent first. Only returns AI-analyzed meetings. Without filters, returns meetings across all advisors in the organization. Pass advisor_id='{user_id}' to show only the current advisor's meetings (use when user says 'my meetings'). Pass a specific advisor ID to query another advisor's meetings — use users_list_users to look up available user IDs. Optionally filter by contact_id. Optionally filter by date range using from_date and to_date (ISO 8601 format, e.g. '2024-01-15' or '2024-01-15T14:00:00'). to_date must not be in the future. Both are optional, omit to_date to get all past meetings from a given date, or omit from_date to get all past meetings up to a given date. Use room_id and session_id from results to fetch transcripts or AI results. Some fields in the response may be redacted for privacy and replaced with placeholder tokens (e.g. <PERSON_NAME>, <EMAIL_ADDRESS>). Treat these as intentionally withheld data — never guess, infer, or fabricate the real value, and don't invent your own stand-in for it either (no brackets, tokens, or notes marking that something was hidden), unless the user explicitly asks to see them. Just write around the missing detail, the same way you would if it had never come up.

List upcoming meetings (from now onward), sorted by soonest first. Without filters, returns meetings across all advisors in the organization. Pass advisor_id='{user_id}' to show only the current advisor's meetings (use when user says 'my meetings'). Pass a specific advisor ID to query another advisor's meetings — use users_list_users to look up available user IDs. Optionally filter by contact_id. Optionally filter by date range using from_date and to_date (ISO 8601 format, e.g. '2025-06-15' or '2025-06-15T14:00:00'). from_date must not be in the past. Both are optional, omit to_date to get all upcoming meetings from a given date, or omit from_date to get all upcoming meetings up to a given date. Use room_id and session_id from results to fetch transcripts or AI results. Some fields in the response may be redacted for privacy and replaced with placeholder tokens (e.g. <PERSON_NAME>, <EMAIL_ADDRESS>). Treat these as intentionally withheld data — never guess, infer, or fabricate the real value, and don't invent your own stand-in for it either (no brackets, tokens, or notes marking that something was hidden), unless the user explicitly asks to see them. Just write around the missing detail, the same way you would if it had never come up.

Get full details for a specific user by ID. Returns name, email, phone, status, and roles. Use an id from a prior list_users response. Some fields in the response may be redacted for privacy and replaced with placeholder tokens (e.g. <PERSON_NAME>, <EMAIL_ADDRESS>). Treat these as intentionally withheld data — never guess, infer, or fabricate the real value, and don't invent your own stand-in for it either (no brackets, tokens, or notes marking that something was hidden), unless the user explicitly asks to see them. Just write around the missing detail, the same way you would if it had never come up.

List users in the current account. Returns user IDs, names, emails, phone numbers, status, and roles. Without filters, returns all users in the account. Use this tool to look up user IDs when another operation requires one — for example, to scope results to a specific advisor or user. Optionally filter by name (partial match on first or last name), emails (exact match), specific IDs, or status (e.g. 'active', 'inactive'). Results are paginated; use page (0-based) and size to control pagination. Some fields in the response may be redacted for privacy and replaced with placeholder tokens (e.g. <PERSON_NAME>, <EMAIL_ADDRESS>). Treat these as intentionally withheld data — never guess, infer, or fabricate the real value, and don't invent your own stand-in for it either (no brackets, tokens, or notes marking that something was hidden), unless the user explicitly asks to see them. Just write around the missing detail, the same way you would if it had never come up.