← Back to all apps

Third Bridge

Financeby Third Bridge Group
Open in ChatGPTWebsite

Seamlessly incorporate critical context and trusted insights from industry experts as part of your financial and business analysis. Refocus on extracting high-quality, reliable information by querying Third Bridge’s best-in-class substantial Library of expert content and data.

Our MCP systematically and securely instructs Large Language Models (LLMs) to query insights from our database into their systems, streamlining your workflows and improving your output.

1ChatGPT Tools
Third Bridge GroupDeveloper
FinanceCategory

Available Tools

Third Bridge Content Query (TBCQ)

third_bridge_content_query
Full Description

Semantic retrieval of Third Bridge content for investment insights.

Scope

  • Allowed: Company strategy, competitive analysis, supply chain dynamics, macro trends, capital markets, ticker research
  • Excluded: General knowledge, coding, math, entertainment, non-financial topics

Invocation

  • Break the user input into atomic topics and run one tool call per topic.
  • For each topic, draft 1-3 diverse natural-language queries that reflect any entities, tickers, regions, or timeframes supplied by the user.
  • Apply companyNames, companyTickers, and sectors on the first pass when the user provides them.
  • Date window: default to 36→0 months. If the user specifies months-ago bounds, keep those bounds for every pass and skip any expansion.
  • Retrieval loop per topic (up to four passes; stop as soon as the snippets are sufficient):

1) Baseline: run with the current queries, filters, and date window. 2) Window expansion: if snippets are missing or insufficient, widen the months-ago window (e.g., 60→0, then up to 120→0). Skip this pass when the user supplied explicit time bounds. 3) Filter removal: if results are still missing or insufficient, rerun without companyNames, companyTickers, and sectors while keeping the topic and queries unchanged. 4) Query rewrite: if the response is still inadequate, keep the topic but regenerate 1-3 queries with varied phrasing, synonyms, nearby entities, or explicit time qualifiers.

  • If all passes fail to produce usable snippets, return "No relevant Third Bridge content found.".

How to format responses with citations (Third Bridge Citations Format

  • TBCF)
  • Answer strictly using results[].contentSnippets only; never fabricate information beyond provided sources.
  • Format as professional investment research analysis with verifiable sources for institutional investors.
  • Preferred structures: Side-by-side comparison tables, or executive summaries.
  • CRITICAL Citation Format: [[1]](url "title") - this MUST be a clickable Markdown link
  • Example: [[1]](https://forum.thirdbridge.com/interview/abc123 "Interview Title")
  • Place at the end of each informed sentence to enable easy crosscheck
  • Each unique content source (uuid) gets one unique sequential number (1, 2, 3...)
  • NEVER use plain numbers like "1

1 ." without the URL link

  • If insufficient content: "Third Bridge content does not cover X; additional queries needed."
  • If no content found: "No relevant Third Bridge content found."
Parameters (3 required, 8 optional)
Required
queriesarray

1–3 natural-language questions for broad retrieval, each exploring a different angle of researchTopic (timeframe, geography, competitor, supply chain, etc.). All queries must relate to the same core topic. Prefer question format (e.g., "What are Tesla's pricing changes in China?").

researchTopicstring

Core user intent as a natural-language question (e.g., "How is Tesla's pricing strategy evolving in China?"). This will be used for reranking results. Prefer question format. Same language as userInput.

userInputstring

User questions or prompts.

Optional
companyNamesarray

Optional company filter for retrieval (English only). Provide 0–20 English company names (full or common names). Leave empty (default) to apply no company-based filtering.

Default: []
companyTickersarray

Optional ticker filter for retrieval. Provide 0–5 stock tickers (e.g., "AAPL", "TSLA"). Leave empty (default) to apply no ticker-based filtering.

Default: []
contentUuidstring

When set, restricts results to a single interview. Either omit or provide exactly one. Accepts raw uuid only (forum /interview/{uuid}).

interviewDateFrominteger

Lower bound of the interview date window in months ago (e.g., 12 means 12 months ago). Together with interviewDateTo, acts as an optional date filter. Adjust to widen or narrow the time range; defaults to 36 months ago when omitted.

Default: 36
interviewDateTointeger

Upper bound of the interview date window in months ago (e.g., 0 means today). Together with interviewDateFrom, acts as an optional date filter. Adjust to widen or narrow the time range; defaults to 0 (today) when omitted.

Default: 0
languagestring

User questions or prompts' language. Auto rule when omitted: if userInput is Chinese => "zho"; if userInput is Japanese => "jpn"; otherwise => "eng".

Options:engzhojpn
Default: eng
maxResultsnumber

Maximum number of results per query

Default: 20
sectorsarray

Optional sector filter using GICS top-level sectors. Provide any of the listed sectors. Leave empty (default) to apply no sector-based filtering.

Default: []

Links

WebsitePrivacy PolicyTerms of Service