mcpField
F
filingrail-mcp (SEC EDGAR data)

filingrail-mcp (SEC EDGAR data)

Local
rapidapi.comAI & Agents0Updated 3w ago

About filingrail-mcp (SEC EDGAR data)

filingrail-mcp is an MCP server that brings SEC EDGAR data to clients such as Claude Desktop, Cursor, and Continue. It structures filings and company data as JSON, with SEC source filing links for financial results.

Key features

  1. 01Search SEC companies
  2. 02Retrieve normalized financials and history
  3. 03Query filings, insider trades, and 8-K events
  4. 04Look up institutional 13F holdings
  5. 05Trace financial data to source filings

Use cases

  1. 01Review recent company financial metrics
  2. 02Track insider purchases and sales
  3. 03Check material company 8-K events
  4. 04Inspect quarterly institutional holdings

MCP support

The MCP server exposes SEC EDGAR tools for company search, financials, filings, insider trades, 8-K events, and 13F holdings. Clients launch filingrail-mcp over stdio and provide RAPIDAPI_KEY.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config
{
"mcpServers": {
"filingrail": {
"command": "uvx",
"args": [
"filingrail-mcp"
],
"env": {
"RAPIDAPI_KEY": "<RAPIDAPI_KEY>"
}
}
}
}
Run commanduvx filingrail-mcp
Environment variables to set
RAPIDAPI_KEYSecretYour RapidAPI key for the Filingrail API. Get one free (50 calls/day, no card) at https://rapidapi.com/hudson-enterprises-llc-hudson-enterprises-llc-default/api/filingrail

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages
pypifilingrail-mcp0.1.3

Tools

8

Saved tool list · updated 3w ago

ParameterTypeRequiredDescription
qstringYes
limitintegerNo

Details

From the project README

filingrail-mcp — SEC EDGAR filings inside Claude, Cursor, and any MCP agent

Give your AI agent live SEC EDGAR data: company financials, insider trades, 8-K events, 13F holdings, and the raw filings stream — all normalized to clean JSON, every number traceable back to its sec.gov source filing.

One pip install, one API key, and your agent can answer "What was Apple's cash position last quarter?" or "List Berkshire Hathaway's top 10 holdings as of 2026-Q1" — against real filings, not its training data.

pip install filingrail-mcp

Live on PyPI: pypi.org/project/filingrail-mcp · MIT licensed · Python 3.10+

Why this exists

Every number a public company reports to the SEC is free and open on EDGAR. The bottleneck isn't access — it's the XBRL parser, the tag-drift normalizer, the rate-limit governor, and the dedup layer you'd have to write before you get the one figure you actually wanted. filingrail-mcp hands your agent the finished JSON instead.

  • Source-traceable. Every financial response carries the originating source_filing_url. Click it, see the exact 10-K / 10-Q on sec.gov. No "trust us."
  • Normalized. revenue is revenue, whether the filer tagged it us-gaap:Revenues, SalesRevenueNet, or RevenueFromContractWithCustomerExcludingAssessedTax. 1.8M+ XBRL rows back to 2006.
  • Zero plumbing. No EDGAR User-Agent config, no local parser, no rate-limit governor. The API handles all of it.

Get a key (free tier, no card)

Filingrail is a REST API on RapidAPI. The free tier is 50 calls/day, no credit card:

→ Subscribe to Filingrail on RapidAPI and copy your X-RapidAPI-Key.

The end user brings their own key — set it as RAPIDAPI_KEY in your MCP client config (below).

Configure your agent

Claude Desktop

Edit %APPDATA%/Claude/claude_desktop_config.json (Windows) or ~/Library/Application Support/Claude/claude_desktop_config.json (macOS):

{
  "mcpServers": {
    "filingrail": {
      "command": "filingrail-mcp",
      "env": { "RAPIDAPI_KEY": "your-rapidapi-key-here" }
    }
  }
}

Restart Claude Desktop. The Filingrail tools (search_companies, get_financials, get_insider_trades, …) appear in the tool list. Then just ask a question in plain English.

Cursor / Continue / any other MCP client

Point the client at filingrail-mcp as the stdio command and pass RAPIDAPI_KEY as an environment variable. Same shape as above.

If your client only accepts a raw command, use:

{ "command": "filingrail-mcp", "env": { "RAPIDAPI_KEY": "your-rapidapi-key-here" } }

Tools exposed

ToolWhat it does
search_companies(q, limit)Resolve ticker / CIK / name fragment to a company (8,000+ SEC issuers, fuzzy match)
get_recent_filings(form_type, cik, limit)Recent filings firehose — filter by form (10-K, 8-K, 4, 13F-HR) or CIK
get_financials(ticker_or_cik, limit)Latest 10-K / 10-Q normalized line items + source_filing_url
get_financials_history(ticker_or_cik, period, limit, …)Historical quarterly or annual series
get_insider_trades(ticker_or_cik, since, limit)Form 4 transactions — who bought/sold, when, how much
get_8k_events(ticker_or_cik, since, limit)8-K material events with SEC item codes (5.02, 2.01, …)
get_13f_holdings(institution_cik, quarter, limit)Institutional holdings (e.g. Berkshire = CIK 1067983)
health()API liveness check

Try it

After installing and configuring, ask your agent:

  • "What was Apple's revenue and operating income last quarter, and link me the filing."
  • "Show me insider sells at NVIDIA in the last 30 days."
  • "What 8-K events did Tesla file this month?"
  • "List Berkshire Hathaway's 10 largest 13F positions for 2026-Q1."

What it is not

Honest caveats:

  • Not real-time. New filings appear within ~24h of EDGAR acceptance (fair-access policy).
  • Not Bloomberg. No intraday prices, options chains, non-US issuers, or analyst estimates.
  • Structured SEC data, traced to source — that's the whole product.

Pricing

TierPriceCalls/month
Free$01,500 (50/day)
Pro$9/mo5,000
Ultra$49/mo50,000
Mega$199/mo500,000

Keys and billing are handled by RapidAPI. Get yours →

Links

License

MIT © Hudson Enterprises LLC

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.