Plug audit-grade SEC data into your AI workflows.
First-source SEC, transcripts, IR — every fact clickable to the source filing. MCP-native, free tier, no card. The accessible US-SEC data layer for the builders the enterprise vendors price out.
Plug into your AI in 30 seconds →Does it actually help? We measured it.
Plug audit-grade SEC data into your AI workflows.
Same model, same questions — with MetricDuck the agent reaches the answer in roughly a third of the steps, at about half the cost. On the vals.ai Finance-Agent v2 set (n=27): with MetricDuck $0.77/query & 28 turns vs $1.65 & 88 without — accuracy delta +0.037 (final pending).
The data layer makes finance agents better — a category now validated by well-funded players. The enterprise options are paid, sales-gated, and built for institutions. MetricDuck is the free, self-serve, US-SEC-focused one you can verify yourself before connecting.
What is the MCP server?
MetricDuck's implementation of the Model Context Protocol — a single HTTP endpoint that lets any MCP client (Claude Code, Claude Desktop, Cursor, Windsurf, or a custom agent) query MetricDuck's financial database in natural language. Transport is Streamable HTTP; auth is automatic via OAuth 2.1, or an API key for headless agents and CI.
18 tools across financials, filing-signal screening, peer comparison, and full-text SEC search — see the tool reference.
Data sources & provenance
MetricDuck is multi-source. Most data originates in SEC filings, but not all of it — here is exactly what comes from where.
Financial statements & XBRL facts
Parsed directly from SEC EDGAR filings — 10-K, 10-Q, 8-K, DEF 14A, and Foreign Private Issuer forms 20-F / 6-K / 40-F. Updated daily. Every number deep-links to the filing it came from.
Derived & computed metrics
323+ metrics computed from the EDGAR-sourced fundamentals (formulas plus dimensional variants) — not third-party estimates or consensus. The inputs to each metric trace back to XBRL facts.
Qualitative filing signals
Tone, guidance, risk and IR-event signals are derived from the filing text plus earnings-call transcripts and IR disclosures — so signal coverage is broader than SEC filings alone.
Market prices & valuation multiples
Daily end-of-day prices come from a market-data feed and power the valuation multiples (P/E, EV/EBITDA, P/B). Limitations: end-of-day only — no intraday or real-time quotes, no options, no analyst estimates.
For Foreign Private Issuers, returned filing-text carries an accounting_standard field (US-GAAP or IFRS) so agents can branch when comparing IFRS-vs-GAAP ratios.
Try asking your agent
Verify a claim against the source
“Did Costco's CEO actually say tariffs won't affect Kirkland Signature pricing? Cite the source.”
Compare peers and surface anomalies
“Compare NVDA vs AMD vs Broadcom on margins, ROIC, valuation. Surface any anomalies.”
Get the setup
“Show me the MetricDuck MCP config for Claude Desktop.”
See it work
Worked end-to-end examples — real prompts, real SEC data, with the manual work they replace.
Connect your client
Claude Desktop
Add to claude_desktop_config.json:
{
"mcpServers": {
"metricduck": {
"command": "npx",
"args": ["mcp-remote", "https://mcp.metricduck.com/mcp"]
}
}
}Cursor
Add to ~/.cursor/mcp.json (or via Settings → MCP):
{
"mcpServers": {
"metricduck": {
"type": "http",
"url": "https://mcp.metricduck.com/mcp"
}
}
}ChatGPT (Developer Mode)
Settings → Connectors → New connector → paste this URL:
https://mcp.metricduck.com/mcpOther clients (Claude Code, Windsurf, headless agents with API keys) — full setup guide.
Free tier, no card.