SEC data for AI agents: how the options compare
ko.io vs sec-edgar-mcp
The free SEC EDGAR APIs and MCP servers, compared. Open-source tools win some jobs; here is where each one fits.
When the open-source tools are the better choice
Use sec-edgar-mcp or raw EDGAR when...
- You need one company's raw filing text (10-K sections, exact XBRL concepts).
- You want an open-source tool you can self-host, with no vendor.
- You want to store, process and repeatedly query downloaded data yourself; fetching data from the SEC remains subject to its access rules.
Check each project's documentation, license and current maintenance status.
Use ko.io when...
- You ask questions that span the whole market, not one filing.
- You want SEC data alongside congressional trades and macro series.
- Your agent runs in a hosted AI client and needs a remote MCP server.
- You want structured, precomputed results without running the data pipeline yourself.
What each one can actually answer
- Get one company's raw 10-K / 8-K / 13F
- DIY EDGARYesko.ioYes (filings gateway)
- Which institutions increased their reported NVDA holdings?
- DIY EDGARNo retrieve and compare every filer's 13F yourselfko.ioYes cross-institution query
- Berkshire's portfolio, 52 quarters back
- DIY EDGARretrieve and normalize historical filings, or use the tool's existing history and cacheko.ioYes precomputed (Pro)
- A manager's filing entities as one portfolio
- DIY EDGARNoko.ioYes family consolidation
- Congress trades and macro series
- DIY EDGARNo not on EDGARko.ioYes
- Works in hosted AI clients
- DIY EDGARdepends on the selected tool's transport and deploymentko.ioYes hosted MCP
- Setup
- DIY EDGARinstall and configure the selected tool as documentedko.ioadd the remote server in a compatible client; see the client-specific setup
- Open source
- DIY EDGARYesko.ioSDKs open, data layer hosted
- Price
- DIY EDGARfree (SEC rate limits)ko.ioFree: 200 calls a day, latest quarter. Pro: $29/mo
Free SEC EDGAR API alternatives
Many tools in this space wrap raw EDGAR access. ko.io is a hosted analytics layer across filers, quarters and sources, built for AI agents. Project descriptions summarize their public documentation; check it for current features.
- sec-edgar-mcpOpen-source MCP server
- Best for: Single-company raw filings and XBRL, self-hosted
- edgartools / raw EDGARPython library
- Best for: Building your own EDGAR parser, with full control
- Financial Datasets MCPHosted MCP
- Best for: Fundamentals and price data by ticker
- ko.ioHosted API + MCP
- Best for: Analytics across the market: 13F across all filers, congressional trades, macro, bitcoin ETF holders and manager families, usable from compatible hosted AI clients
Why not scrape EDGAR yourself?
You can. Here is the work it involves, and what ko.io provides instead.
- Time to first clean row
- Do it yourself: Use an existing parser or implement extraction for the fields you need
- With ko.io: One API call
- Normalization and dedupe
- Do it yourself: You build it (share classes, CIK merges)
- With ko.io: Consolidated records where the dataset supports it
- Provenance
- Do it yourself: You link every number back to its filing yourself
- With ko.io: Parsed from public filings; the filings endpoints return the source documents
- Serving
- Do it yourself: EDGAR rate limits plus your own infrastructure
- With ko.io: Hosted REST API and MCP server, served from two regions
- AI access
- Do it yourself: Configure a compatible MCP server or another integration
- With ko.io: A hosted MCP server you add in any compatible client
- Commercial rights
- Do it yourself: Yours to work out
- With ko.io: Use rights per plan, set out in the Terms
Try it without signing up
MCP (Claude, Cursor, any client)
claude mcp add ko-sec-data --transport http https://mcp.ko.io/mcpREST (keyless demo)
curl "https://api.ko.io/api/v1/institutions?search=berkshire&demo=true"Python
pip install ko-edgarCommon questions
- Is there a free SEC EDGAR API?
- Yes, several. SEC EDGAR itself is free (rate-limited); open-source tools such as sec-edgar-mcp and edgartools wrap it for local use; and ko.io has a Free plan (200 calls a day, most recent quarter) over REST and MCP. Which is best depends on whether you need one company's raw filings (use the open-source tools) or analytics across the market, such as "which institutions increased their NVDA holdings" (use ko.io).
- What is the best SEC data MCP server for AI agents?
- For a single company's raw filings, sec-edgar-mcp is a good open-source choice. For questions that span the whole market (13F changes across all institutions, congressional trades, macro series, bitcoin ETF holders), or for hosted AI clients, ko.io's hosted MCP server is built for the job: precomputed and parsed from the public filings. In Claude Code, connect with one command:
claude mcp add ko-sec-data --transport http https://mcp.ko.io/mcp - What is a good sec-edgar-mcp alternative?
- If you need aggregation and breadth rather than one company's raw filings, consider the hosted ko.io. Results across institutions and quarters are precomputed, so a query does not re-parse historical filings; it adds data beyond EDGAR (congressional trades, macro series); and it runs as a remote MCP server, usable in hosted AI clients that support remote MCP.