
U.S. Government Open Data
io.github.lzingav2026.10.8更新于 Oct 8, 2026
Open data from 40+ U.S. government APIs: economy, spending, Congress, health, energy, disasters
概览
让助手通过 300 多个工具查询 40 多个美国政府开放数据 API,涵盖经济、健康、立法、支出、能源等领域。
- 功能
- 提供 300 多个工具,接入 40 多个美国官方及国际 API,覆盖经济指标、财政与采购数据、国会与游说、健康与安全、环境与天气、教育、专利和人口统计。内置指引鼓励在一次回答中交叉引用多个机构的数据,每个结果都会列出其上游 API 请求。WASM 沙箱代码模式可执行 JavaScript 以精简大型响应,选择性加载或发现模式可减少暴露的工具数量。
- 适用场景
- 适合助手需要权威美国公共数据来回答研究、报道、政策或分析类问题,尤其是需要把经济指标与立法或健康数据结合时。若只是单次查询、一个 API 调用即可解决,则没有必要安装。
- 运行要求
- 通过 npx 以 stdio 在本地运行,需要 Node.js 22 或更高版本。许多 API 无需密钥;其他 API 需要免费密钥,以环境变量形式提供,例如 FRED_API_KEY、DATA_GOV_API_KEY、CENSUS_API_KEY、BLS_API_KEY、EIA_API_KEY、NOAA_API_KEY 等。AQS_EMAIL、SEC_CONTACT_EMAIL 和 NWS_USER_AGENT 用于标识调用者。需要能访问这些政府 API 的网络。
安装
在 SourceWeft 中
- 打开 控制台中的 U.S. Government Open Data,将其添加到工作区。
- 为需要使用其工具的对话启用该服务。
Desktop only,通过 STDIO。 STDIO 服务会启动本地进程,因此需要 SourceWeft 桌面宿主。
其他 MCP 客户端
参照 仓库 中的启动说明。
README
US Government Open Data MCP
MCP Server + TypeScript SDK for 40+ U.S. Government APIs
[npm version] [npm downloads] [License: MIT]
300+ tools covering economic, fiscal, health, education, energy, environment, lobbying, housing, patents, safety, banking, consumer protection, workplace safety, transportation, seismic, clinical trials, pharma payments, research funding, procurement, and legislative data.
20+ APIs require no key · The rest use free keys that take under a minute to get
Features
- 300+ tools across 40+ government APIs — economic, health, legislative, financial, environmental, and more
- Cross-referencing — built-in instructions guide the LLM to combine data from multiple agencies (e.g., FDA adverse events + lobbying spend + campaign contributions)
- Code mode — WASM-sandboxed JavaScript execution reduces context window usage by 98-100% for large responses
- Selective loading — load only the modules or domains you need:
--modules fred,treasury,congressor--domains economy,health - Discovery mode —
--tool-mode discoverylists 4 tools instead of 350; the model searches for the one it needs - Sources on every result —
meta.sourceslists the upstream API requests behind each answer - Dual transport — stdio for desktop clients, HTTP Stream for web/remote (bearer-token auth off loopback)
- TypeScript SDK — every API is importable as a standalone typed client, no MCP required
- Disk-backed caching — responses cached to disk, survives restarts
- Rate limiting + retry — token-bucket rate limiter with exponential backoff on 429/503
Quick Start
MCP Server
Requires Node.js 22 or later.
Add to .vscode/mcp.json for VS Code / Copilot:
Add to claude_desktop_config.json for Claude Desktop:
Example Prompts
Once connected, ask your AI assistant natural language questions:
Economic: "What's the current state of the U.S. economy? Show me GDP, unemployment, inflation, and interest rates."
Health: "Show me the adverse event profile for Ozempic including clinical trials, FDA reports, and pharma payments to doctors."
Legislative: "What happened with the Inflation Reduction Act? Who sponsored it, how did the vote break down by party?"
Follow the money: "Which banking PACs gave money to members of the Senate Banking Committee, and how did those members vote on banking deregulation?"
Cross-reference: "How has federal spending on healthcare changed over the last 5 years, and what health outcomes has it produced?"
TypeScript SDK
No MCP server required. All functions include caching, retry, and rate limiting.
Documentation
Full documentation at lzinga.github.io/us-gov-open-data-mcp
Data Sources
Disclaimer
This project integrates a significant number of government APIs, many of which have large, complex, or inconsistently documented schemas. AI is used as a tool throughout this project to help parse API documentation, generate type definitions, and scaffold tool implementations — making it possible to cover this much surface area and get people access to government data faster than would otherwise be feasible. While every effort has been made to ensure accuracy, some endpoints may return unexpected results, have incomplete parameter coverage, or behave differently than documented.
This is a community-driven effort — if you find something that's broken or could be improved, please open an issue or submit a PR. Contributions that fix edge cases, improve schema accuracy, or expand coverage are especially welcome. The goal is to make U.S. government data as accessible and reliable as possible, together.
All data is sourced from official U.S. government and international APIs — the server does not generate, modify, or editorialize any data.
License
MIT
来源:README.md,提交 06df35c
工具
0版本历史
1- v2026.10.8最新Oct 8, 2026


