gov-data-mcp

95 US government open-data tools, as one MCP server.

EPA, FEMA, USGS, NOAA, FAA, USACE, FDIC, HUD, NRCS, HRSA, CMS, county assessor rolls and state licensing boards — all reachable as agent-callable tools, all reading directly from official government APIs and bulk files. No scraping, no HTML parsing, no rate-limit roulette.

mcp-name: io.github.malonestar/gov-data-mcp

npx gov-data-mcp

Install

You need a free Apify API token from console.apify.com/settings/integrations. Apify's free tier includes monthly platform credit that covers evaluating every tool here.

Claude Desktop / Claude Code / Cursor

{
  "mcpServers": {
    "gov-data": {
      "command": "npx",
      "args": ["-y", "gov-data-mcp"],
      "env": { "APIFY_TOKEN": "apify_api_..." }
    }
  }
}

Claude Desktop reads claude_desktop_config.json; Claude Code reads .mcp.json in your project; Cursor reads .cursor/mcp.json. The block is identical in all three.

What you get

12 dedicated tools for the highest-traffic questions, callable directly:

Tool	Answers
site-due-diligence-bundle	20-layer go / caution / no-go verdict for one coordinate
epa-contaminated-site-screener	Phase I ESA database search at ASTM E1527-21 distances
faa-drone-airspace-checker	Part 107 LAANC and airspace verdicts, batched
hifld-grid-proximity-screener	Nearest transmission line, substation, serving utility, ISO/RTO
interconnection-queue-tracker	Generator interconnection queues across 7 ISOs
fdic-ncua-health-rollup	Bank and credit-union health with real peer cohorts
fema-nri-county-risk-profile	County and census-tract natural-hazard risk
fws-wetlands-proximity-screener	NWI wetlands within a radius, with decode columns
nhd-surface-water-404-screener	Clean Water Act §404 surface-water screen
epa-drinking-water-quality-screener	SDWA violations, lead, PFAS occurrence
parcel-owner-lookup	Assessor-roll owner of record for an address
license-verifier	19 professional licensing boards across 9 states + OIG exclusions

Plus three tools that reach the other 83:

• search_gov_data_tools — find a tool by keyword, agency or topic
• describe_gov_data_tool — full input schema for any tool in the catalog
• run_gov_data_tool — run any tool in the catalog

The catalog is bundled, so discovery costs nothing. Ask your agent "what government data tools do you have for flood risk?" and it will search all 95.

Example prompts

Screen 1200 Broadway, Denver CO for environmental risk under ASTM E1527-21 and tell me which findings fall inside the standard's search distance.

I'm siting a 40 MW solar project at 41.88, -93.10. Check grid proximity, prime farmland, wetlands, critical habitat and the interconnection queue, then tell me what would kill the project.

Can I fly a Part 107 mission at these six coordinates, and which ones need a DroneZone authorization rather than LAANC?

Which Texas banks show the 2006 CRE-concentration guidance flagged on both prongs?

How it works

Each tool is a published Apify Actor that this server invokes through the Apify API. The server starts the run, waits for a terminal state, and returns the rows.

Billing is to your own Apify account at each actor's published pay-per-result rate, listed on its Store page. Free-tier credit covers evaluation. A run that fails bills nothing beyond a fractional actor-start charge.

Every result carries the run_id and a console.apify.com URL, so any claim an agent makes from this server can be traced back to the exact run that produced it.

On honest answers

These actors are built around one rule: a failure must never be presentable as "nothing was found." That distinction matters most in exactly the cases people use this for — telling a buyer a property is clear of contamination, telling a pilot an airspace is uncontrolled, telling a lender a borrower is unlicensed.

So this server:

• reports run_status on every call, and never attaches a rows key to a run that did not succeed
• distinguishes "the run SUCCEEDED and the source genuinely matched nothing" from "the run failed" in the response text, explicitly
• retries transient Apify 429/5xx, then fails loudly saying the platform failed and no conclusion should be drawn about the government source
• treats an unknown tool name as a catalog miss, never as an empty result

The underlying actors carry the same discipline: per-source status on every row, null meaning not checked rather than checked and negative, live drift assertions that fail the run when an upstream silently truncates, and pagination guards on sources that answer HTTP 200 with a partial payload. Each actor's Store page documents the specific traps it defends against.

Coverage

Environmental & contamination · flood, wildfire, seismic, landslide, karst & sea-level rise · wetlands, critical habitat, protected lands & historic places · energy siting, grid, pipelines & interconnection · farmland, soils & water rights · banking, credit unions & fair lending · securities, auditors & pension plans · professional licensing & exclusion screening · real-estate leads, parcels, deeds & distress signals · drone airspace, bridges, tunnels & dams.

Run search_gov_data_tools with any of those terms for the full list, or browse apify.com/malonestar.

Development

npm install
npm test              # offline suite, no network
node tools/mutate.cjs # re-injects known defects, asserts the suite catches them
npm run catalog -- <apify-token>   # regenerate src/catalog.json from the live API

License

MIT