filingrail-mcp
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.
README
<!-- mcp-name: io.github.adamhudson777/filingrail-mcp -->
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.
revenueisrevenue, whether the filer tagged itus-gaap:Revenues,SalesRevenueNet, orRevenueFromContractWithCustomerExcludingAssessedTax. 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
| Tool | What 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
| Tier | Price | Calls/month |
|---|---|---|
| Free | $0 | 1,500 (50/day) |
| Pro | $9/mo | 5,000 |
| Ultra | $49/mo | 50,000 |
| Mega | $199/mo | 500,000 |
Keys and billing are handled by RapidAPI. Get yours →
Links
- Get an API key (RapidAPI listing)
- Filingrail homepage
- Python SDK (sync + async) —
pip install filingrail - API status page
- Questions / issues:
support@hudsonenterprisesllc.com
License
MIT © Hudson Enterprises LLC
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。