the13f-mcp

mcp-name: com.the13f/mcp

Model Context Protocol server for the13f. Brings institutional 13F intelligence into Claude Desktop, Cursor, VS Code + Continue, and any MCP-compatible host — no HTTP code required on your side.

Important limitations

Research data only. Sourced from publicly disclosed SEC Form 13F filings, which lag quarter-end by up to 45 days. Past institutional positioning does not predict future performance. Nothing here constitutes investment advice. Every tool's response includes a disclaimer field; the host LLM should surface it alongside any analysis it produces.

Install

uvx the13f-mcp

Or pip install the13f-mcp if you prefer to manage Python environments yourself. Python 3.11+ required.

Get a free API key

Free, no payment, no credit card:

1. Visit https://the13f.com/integrations/
2. Enter your email and solve the Cloudflare Turnstile challenge
3. Copy your pf13f_ key — shown once
4. Paste it into your MCP host's config as THE13F_API_KEY (see below)
5. Restart the host

Free tier: 100 read calls per day; quota rolls at UTC midnight. Signal and report tools arrive in later minor versions when paid tiers launch.

Configure your MCP host

Claude Desktop

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

{
  "mcpServers": {
    "the13f": {
      "command": "uvx",
      "args": ["the13f-mcp"],
      "env": {
        "THE13F_API_KEY": "pf13f_..."
      }
    }
  }
}

Restart Claude Desktop. The13f tools should appear in the MCP tool list.

Cursor

Settings → MCP → Add Server. Paste the same JSON block into the "Custom" field.

VS Code with Continue

Settings → experimental.modelContextProtocolServers → add a stdio transport running uvx the13f-mcp with THE13F_API_KEY in env. The key reveal page at https://the13f.com/integrations/signup-success shows a copy-paste-ready snippet with the key already filled in.

Available tools — v0.2.0 (Read tier, free)

Planned tools — later paid releases

The free-tier Read tools are enough to explore the data and build prompts against it. Signal and report tools deliberately wait on the paid-tier billing plumbing.

Example prompts

"Pull Berkshire Hathaway's top 20 positions as of Q4 2025 and summarize what grew the most quarter-over-quarter."

"I own AAPL, MSFT, and NVDA equal-weight. Find the 5 institutional managers whose portfolios most closely resemble mine."

"What sector did 13F filers most aggressively reduce last quarter? Show me the top three managers leading the reduction."

"Which institutional managers hold NVDA this quarter, and who newly bought or exited it versus last quarter?"

"What did Renaissance buy, sell, add, and trim last quarter, and where does its book overlap most with Two Sigma's?"

The MCP host's LLM picks the right tool, fills in arguments, surfaces the disclaimer, and returns the result as structured JSON.

Environment variables

Source

Public source: https://github.com/pickelfintech/the13f-mcp. This is the snapshot that PyPI, Glama.ai, and the MCP community list point at. The GitHub repo is a push-mirror of the GitLab one at gitlab.com/pickel-fintech/the13f-mcp; GitLab remains the primary and also carries the release CI. Bug reports and PRs are accepted on either side.

License

MIT — see LICENSE. The hosted api.the13f.com API that this client calls is a separate service. A free API key (100 calls/day) is required; sign up at https://the13f.com/integrations/.

Support

tom@pickelfintech.com — quote the request_id field returned by any failing tool call for fastest triage.

License

MIT. See LICENSE. The MCP server source is MIT-licensed; use of the the13f API itself is governed by the Terms of Service at https://the13f.com/terms.html.