Disclaimer 🚨

EDGAR® and SEC® are trademarks of the U.S. Securities and Exchange Commission. This open-source project is not affiliated with or approved by the U.S. Securities and Exchange Commission.

Introduction 📣

SEC EDGAR MCP is an open-source MCP server that connects AI models to the rich dataset of SEC EDGAR filings. EDGAR (Electronic Data Gathering, Analysis, and Retrieval) is the U.S. SEC's primary system for companies to submit official filings. It contains millions of filings and "increases the efficiency, transparency, and fairness of the securities markets" by providing free public access to corporate financial information. This project makes that trove of public company data accessible to AI assistants (LLMs) for financial research, investment insights, and corporate transparency use cases.

Using the Model Context Protocol (MCP) – an open standard that "enables seamless integration between LLM applications and external data sources and tools" – the SEC EDGAR MCP server exposes a set of tools backed by the official SEC EDGAR REST API. Under the hood, it leverages the Python secedgar SDK (an unofficial wrapper for SEC's API) to fetch data like company filings and financial facts. This means an AI agent can ask questions like "What's the latest 10-K filing for Apple?" or "Show me Tesla's total revenue in 2021" and the MCP server will retrieve the answer directly from EDGAR's official data.

Usage 🚀

Once the SEC EDGAR MCP server is running, you can connect to it with any MCP-compatible client (such as an AI assistant or the MCP CLI tool). The client will discover the available EDGAR tools and can invoke them to get real-time data from SEC filings. For example, an AI assistant could use this server to fetch a company's recent filings or query specific financial metrics without manual web searching.

Demo: Here's a demonstration of an AI assistant using SEC EDGAR MCP to retrieve Apple's latest filings and financial facts (click to view the video):

SEC EDGAR MCP - Demo - Watch Video

In the demo above, the assistant calls the get_submissions tool with Apple's CIK and then uses get_company_concepts to fetch a specific financial concept, showcasing how EDGAR data is retrieved and presented in real-time. 📊

Installation 🛠

Follow these steps to set up and run the SEC EDGAR MCP server:

1. Clone the repository:

git clone https://github.com/yourusername/sec-edgar-mcp.git
cd sec-edgar-mcp

2. Install dependencies: Ensure you have Python 3.9+ installed. Install the required packages, including the MCP framework and secedgar SDK:

pip install mcp secedgar

3. Configure SEC EDGAR API access: The SEC API requires a User-Agent string for all requests. Create a .env file in the project directory and add your user agent info:

SEC_API_USER_AGENT="Your Name (your.email@domain.com)"
PYTHONPATH=/path/to/your/local/cloned/repo/sec-edgar-mcp

This identifies your application to the SEC (replace with your details). The server will load this from the environment to authenticate to EDGAR.

4. Start the MCP server: Launch the server to begin listening for MCP clients. For example:

mcp install sec_edgar_mcp/server.py --env-file .env --name "SEC EDGAR MCP Server" --with secedgar

Once running, the server will register its tools (see below) and await client connections. You should see logs indicating it's ready. 🎉

Now the server is up and running, ready to serve EDGAR data to any MCP client! You can use the MCP CLI or an AI platform (e.g. Claude Desktop) to connect to localhost (or the appropriate transport) and start issuing tool calls.

Tools 🔧

SEC EDGAR MCP exposes several tools (functions) from the SEC EDGAR API. These tools allow retrieval of different types of data:

• Company Submissions – recent filings and metadata for a company (by CIK).
• Company Concept – detailed data for a specific financial concept (XBRL tag) for a company.
• Company Facts – all available financial facts for a company.
• XBRL Frames – aggregated data for a financial concept across companies or time frames.

Each tool is defined with a name, description, and input parameters. AI assistants can invoke them via MCP's JSON-RPC interface. Below is a list of the tools with details and examples of how to call them (click to expand):

{
  "jsonrpc": "2.0",
  "id": 42,
  "method": "tools/call",
  "params": {
    "name": "get_submissions",
    "arguments": { "cik": "0000320193" }
  }
}

{
  "cik": "0000320193",
  "name": "Apple Inc.",
  "tickers": ["AAPL"],
  "exchanges": ["Nasdaq"],
  "filings": {
    "recent": {
      "form": ["10-K", "10-Q", ...],
      "filingDate": ["2023-10-26", "2023-07-27", ...],
      "reportDate": ["2023-09-30", "2023-06-30", ...],
      "primaryDocument": ["aapl-2023-10k.htm", "aapl-2023-q3.htm", ...],
      ... 
    }
  }
}

{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "get_company_concepts",
    "arguments": { 
      "cik": "0000320193",
      "taxonomy": "us-gaap",
      "tag": "AccountsPayableCurrent"
    }
  }
}

{
  "cik": 320193,
  "taxonomy": "us-gaap",
  "tag": "AccountsPayableCurrent",
  "label": "Accounts Payable, Current",
  "description": "The carrying value of accounts payable as of the balance sheet date.",
  "entityName": "Apple Inc.",
  "units": {
    "USD": [
      { "end": "2022-09-24", "val": 64220000000, ... },
      { "end": "2021-09-25", "val": 54763000000, ... },
      ...
    ]
  }
}

{
  "jsonrpc": "2.0",
  "id": 2,
  "method": "tools/call",
  "params": {
    "name": "get_company_facts",
    "arguments": { "cik": "0000320193" }
  }
}

{
  "cik": 320193,
  "entityName": "Apple Inc.",
  "facts": {
    "dei": {
      "EntityCommonStockSharesOutstanding": {
        "label": "Entity Common Stock, Shares Outstanding",
        "units": { "shares": [ ... ] }
      },
      "EntityPublicFloat": {
        "label": "Entity Public Float",
        "units": { "USD": [ ... ] }
      },
      ...
    },
    "us-gaap": {
      "AccountsPayableCurrent": {
        "label": "Accounts Payable, Current",
        "units": { "USD": [ { "end": "2022-09-24", "val": 64220000000 }, ... ] }
      },
      "AccountsReceivableNet": {
        "label": "Accounts Receivable, Net",
        "units": { "USD": [ ... ] }
      },
      ...
    }
  }
}

{
  "jsonrpc": "2.0",
  "id": 3,
  "method": "tools/call",
  "params": {
    "name": "get_xbrl_frames",
    "arguments": { 
      "taxonomy": "us-gaap",
      "tag": "AccountsPayableCurrent",
      "unit": "USD",
      "year": 2019,
      "quarter": 1
    }
  }
}

{
  "taxonomy": "us-gaap",
  "tag": "AccountsPayableCurrent",
  "uom": "USD",
  "ccp": "CY2019Q1I",  <!-- Calendar Year 2019, Q1, Instantaneous -->
  "data": [
    {
      "cik": 1555538,
      "entityName": "SUNCOKE ENERGY PARTNERS, L.P.",
      "end": "2019-03-31",
      "val": 78300000
    },
    {
      "cik": 11199,
      "entityName": "BEMIS CO INC",
      "end": "2019-03-31",
      "val": 465700000
    },
    ... (thousands more data points) ...
  ]
}

Note: The JSON structures above are directly returned from the SEC EDGAR API via the secedgar SDK. The MCP server does not alter the data, so you get the same fields as the official API. All tools require a valid CIK (Central Index Key) for company-specific queries – you can use the SEC's CIK lookup tool if you only know a ticker or name.

Architecture 🏗️

The SEC EDGAR MCP server acts as a middleman between an AI (MCP client) and the SEC's EDGAR backend:

• 🔸 MCP Client: Could be an AI assistant (like Claude or GPT-4 with an MCP plugin) or any app that speaks the MCP protocol. The client sends JSON-RPC requests to invoke tools (e.g. get_submissions) and receives JSON results.

• 🔸 MCP Server (SEC EDGAR MCP): This server (the project you are reading about) defines the EDGAR tools and handles incoming tool requests. When a request comes in, the server uses the secedgar Python SDK to call the SEC EDGAR REST API and then returns the response back over MCP.

• 🔸 SEC EDGAR REST API: The official SEC endpoint (data.sec.gov) that provides EDGAR data in JSON format. The secedgar library communicates with this REST API, abiding by its usage policies (including rate limits and user agent identification).

How it works: The MCP client first queries the server's tool list (discovering functions like get_submissions, etc.). The AI can then decide to call a tool by name with appropriate parameters. The server receives the tools/call request, executes the corresponding EDGAR API call via secedgar, and returns the data. This response is sent back to the AI client in a structured JSON format that the AI can read and incorporate into its answer or reasoning.

In essence, SEC EDGAR MCP bridges the gap between natural language questions and the raw SEC filings data. By adhering to MCP, it standardizes the way AI models can fetch real-world financial data, using officially sourced information for accurate and up-to-date answers.

References 📚

• SEC EDGAR – About EDGAR, SEC.gov (2024). EDGAR is the SEC's database for electronic company filings.

• Model Context Protocol (MCP) – Official documentation and SDKs. ModelContextProtocol.io – An open standard for connecting LLMs to tools.

• SEC EDGAR API Python SDK (secedgar) – An unofficial Python wrapper for SEC's EDGAR REST API. GitHub repo, Documentation.

License ⚖️

This project is licensed under the MIT License. You are free to use, modify, and distribute it. See the LICENSE file for details.

© 2025 Stefano Amorelli – Released under the MIT license. Enjoy! 🎉