MCP Server

Connect AI assistants to SEC filing intelligence via the Model Context Protocol. Tools and recipe prompts across 3 tiers (Free, Pro, Analyst) — company research, financial analysis, signals, real-time monitoring, and click-to-run analysis recipes.

The Model Context Protocol (MCP) lets AI assistants like Claude, ChatGPT, Cursor, and Windsurf call edgar.tools directly — searching companies, pulling financials, analyzing insider trades, and monitoring live filings through natural language. Instead of copy-pasting data or switching between apps, you ask your AI assistant a question and it calls the right edgar.tools tool automatically. Results come back as structured data the assistant can reason about, compare, and synthesize.

The customer-facing landing for the MCP server is the edgar.tools MCP page — visual tier ladder, FAQ, and setup walkthroughs for each AI client. This /docs/mcp section is the technical reference.

How it works#

  1. You ask a question — "What insider trades happened at Tesla in the last 30 days?"
  2. Your AI picks the right tool — Claude automatically calls insider_activity(company='TSLA', days=30)
  3. Structured data comes back — transactions, sentiment analysis, signal detection — ready for the AI to interpret

Tools by tier#

The MCP server's tools/list endpoint advertises only the tools your account's tier includes.

  • Free — Read. search_companies, search_entities, company_filings, company_brief, edgar_recipes, account_status
  • Pro — Track. Adds financial_snapshot, financial_statements, financial_trends, compare_companies, insider_activity, institutional_ownership, material_events, live_filings, search_filings, filing_section, disclosure_search, edgar_notes, fund_profile, adviser_profile, search_funds, search_advisers
  • Analyst — Research. Adds analyze_filing, filing_details, plus the aggregation tools position_intel, peer_facts, portfolio_events, person_intel, account_dossier and the recipe prompts (position_thesis, peer_comparison, portfolio_pulse, earnings_postmortem, filing_red_flags)

See MCP Tiers for the full mapping and how cross-tier calls are handled. See Tools Reference for parameters and return formats. See MCP Prompts for the recipe-prompt catalog.

8 functional categories#

Independent of tier, the tools group into eight categories:

  • Analyze — analyze_filing (paste any SEC filing URL — auto-detects form type)
  • Research — search_companies, company_brief, company_filings, financial_snapshot, financial_statements, financial_trends
  • Compare — compare_companies (side-by-side analysis of 2+ companies)
  • Notes — edgar_notes, disclosure_search, search_filings (drill into narrative text and financial-statement notes)
  • Signals — insider_activity, institutional_ownership, material_events
  • Monitor — live_filings (real-time SEC filing stream)
  • Funds & Advisers — search_funds, fund_profile, search_advisers, adviser_profile
  • Aggregation — position_intel (single-name bundle), peer_facts (peer-comparison rows), portfolio_events (bulk 8-Ks across up to 50 tickers) — designed to be consumed by recipe prompts

Why use MCP instead of the REST API?#

  • Natural language, not endpoints. Ask "compare Apple and Microsoft financials" instead of making 4 API calls and joining the data yourself.
  • AI does the reasoning. Your assistant chains multiple tools, cross-references results, and synthesizes insights automatically.
  • No code required. Connect once in your AI client's settings, then ask questions. No HTTP clients, no authentication code, no parsing.
  • 10–20× cheaper than web search. In a head-to-head benchmark, an agent doing SEC research through this server spent 10–20× fewer tokens and dollars than the same model via web search — and was more accurate. Form-aware responses land in ~2–3K tokens with citations back to the source filing.

Get started#

  • Setup Guide — connect edgar.tools to Claude, ChatGPT, Cursor, or Windsurf in under a minute
  • MCP Tiers — which tools live at which tier, and how cross-tier calls are handled
  • Tools Reference — every tool with parameters, return formats, and tier access
  • Prompts Reference — the 5 Analyst recipe prompts (position_thesis, peer_comparison, portfolio_pulse, earnings_postmortem, filing_red_flags)
  • Example Workflows — prompts and conversations for common analysis tasks
  • MCP landing page — customer-facing visual tier ladder with setup walkthroughs