bigdata_search

Search engine for financial documents, earnings call transcripts, news articles, analyst reports, SEC filings, private uploads, and business content.

Overview

The bigdata_search tool is a search engine for financial documents, earnings call transcripts, news articles, analyst reports, your private uploaded files, SEC filings, and business content from thousands of companies and sources. It returns document chunks with full context, timestamps, and source attribution (URL). Ideal for research, analysis, fact-finding, and gathering company-specific information.

For the best experience, install the Bigdata plugin that includes financial research analyst skills.

Request object

Parameter	Type	Required	Description
request	object	Yes	Search request object containing search_mode ("fast" or "smart") and a query object with text and optional filters.

It supports different query parameters depending on search_mode. For more detail on how fast and smart modes differ, see Fast vs smart search.

fast search_mode

request: You can copy and paste it in the MCP Inspector to test it out.

Search request example (fast mode)

{
    "search_mode": "fast",
    "query": {
      "text": "Netflix's partnership",
      "filters": {
        "document_type": {
          "mode": "INCLUDE",
          "values": [
            {
              "type": "NEWS"
            }
          ]
        }
      }
    }
}

Field	Type	Required	Description
search_mode	string	Yes	"fast" performs a direct semantic and lexical search. "smart" uses an AI agent to interpret the query and perform a more intelligent search.
query.text	string	Yes	Natural-language search query.
filters	object	No	Filter object to narrow results (see below).
query.max_chunks	int	No	Maximum number of chunks to retrieve.

Filters

All filters are optional and go inside request.query.filters.

Filter	Type	Description
reporting_entities	list[string]	Entity IDs of the company that authored or filed the document.
entity	object (any_of / all_of / none_of)	Entity IDs of companies mentioned in the document, regardless of who published it.
document_type	object (mode: INCLUDE/EXCLUDE, values: list)	Filter by document type.
timestamp	object (start / end in ISO 8601)	Filter by publication date range.
category	object (mode: INCLUDE/EXCLUDE, values: list)	Filter by source category: "transcripts", "filings", "news", "research", "my_files".
keyword	object (any_of / all_of / none_of)	Filter by keywords found in documents.
reporting_periods	list of objects (fiscal_year, fiscal_quarter)	Filter by fiscal reporting period.

smart search_mode

request: You can copy and paste it in the MCP Inspector to test it out.

Search request example (smart mode)

{
    "search_mode": "smart",
    "query": {
      "text": "Netflix's partnership in news"
    }
}

Smart mode uses an AI agent to interpret the query. The query object only needs text which describes what to find in natural language (you can mention sources, categories, document types or other filters as part of the query text).

Field	Type	Required	Description
search_mode	string	Yes	"fast" performs a direct semantic and lexical search. "smart" uses an AI agent to interpret the query and perform a more intelligent search.
query.text	string	Yes	Natural-language search query.
query.max_chunks	int	No	Maximum number of chunks to retrieve.

Response object

The response contains a results array, where each element represents a matching document with the following fields:

Field	Type	Description
id	string	The unique identifier of the document, a 32-character hexadecimal string.
headline	string	The title of the document.
timestamp	datetime	The timestamp of the document.
source	object	The source of the document, which contains the name of the source.
chunks	list of objects	The content of the document, returned as relevant text chunks.
url	string	Bigdata App URL that displays the original document. It can be customized to return the original document URL instead (see Returning the original document URL).

Search example response

{
    "results": [
        {
            "id": "0248B5FFA01EA4A5E47EF036E9F79744",
            "headline": "AI for Semiconductor Manufacturing Market Global Insight Services Mar 02, 2026 AI for Semiconductor Manufacturing Market Analysis and Forecast to 2035: Type, Product, Services, Technology, Application, End User, DeploymentAI for Semiconductor Manufacturin",
            "timestamp": "2026-03-03T23:20:59",
            "source": {
                "name": "MarketResearch.com (United States)"
            },
            "chunks": [
                {
                    "cnum": 18,
                    "text": "For instance, in March 2025, TSMC announced a $100 billion investment in U.S.-based semiconductor manufacturing to expand capacity and integrate AI for process optimization and logistics efficiency. Additionally, as reported by Forbes in September 2025, global AI adoption in industrial applications reached 78%, with 88% of executives planning to accelerate deployment underscoring AIs pivotal role in driving innovation, operational excellence, and competitiveness within the semiconductor manufacturing industry.\nGrowing Investment in AI Chip Startups and Advanced Semiconductor Innovation -\nA major trend for the AI for semiconductor manufacturing market is the accelerating investment in AI chip startups and advanced semiconductor innovation.",
                    "relevance": 0.6501799637750973
                },
                {
                    "cnum": 11,
                    "text": "Geographical Overview\nAsia-Pacific region is emerging as a global hub for AI-driven semiconductor manufacturing, fueled by strong policy support, fab expansion, and rising AI compute demand. In Q2 2025, TSMC reported a 38.6% year-over-year revenue surge to US$30 billion, with 74% derived from advanced nodes (3nm, 5nm, 7nm) used in AI and HPC applications, alongside plans for over 15 new fabs globally. In China, Huaweis Ascend 910C chip yield improved from ~20% in 2023 to ~40% in 2024, enhancing domestic AI chip production amid export restrictions.",
                    "relevance": 0.4841535923281182
                }
            ],
            "url": "https://www.marketresearch.com/Global-Insight-Services-v4248/AI-Semiconductor-Manufacturing-44093325/"
        },
       ...
    ]
}

Returning the original document URL

By default, the bigdata_search tool returns the Bigdata App URL of the cited document in the url response parameter. Users who authenticated to the Bigdata MCP with their Bigdata username and password can open that document directly in the Bigdata App.

However, if your proprietary platform uses a single authentication credential or API Key to access the MCP and serve multiple clients, your end users will not have access to the Bigdata App. In that case, you can instruct the MCP server to return the original (public) URL of the document in the url response parameter instead of a link to the Bigdata App.

To do this, your proprietary platform must add the header parameter return-public-url with a value of True:

The url field may be empty if the document is not publicly available. In that case, the client must download the entire document using the document ID.

We also provide a Document Viewer Widget that your proprietary platform can leverage to visualize the entire document.

Usage Monitoring

The tool bigdata_search consumes API Query Units in the same way as the Search Service API. You can monitor the organization`s usage in the Developer Platform > Usage

Prompt examples

Company Brief

Generate a concise company brief with key business insights.

Investment Memo

Build a structured investment memo for a target company.

Earnings Preview

Prepare for upcoming earnings calls with key questions and context.

Earnings Q&A

Analyze earnings call transcripts with targeted questions and answers.