tracts.ai
Log inRequest a sample
MCP Server

County property data, native in your agent.

The tracts MCP server surfaces parcel records as native tools in Claude, Cursor, and any MCP client — the same county-direct data and lineage as the REST API, with no glue code.

Get an API key Read the docs
  • 2 tools
  • OAuth or API key
  • lineage on every record
claude_desktop_config.jsonMCP · Streamable HTTP
{
  "mcpServers": {
    "tracts": {
      "url": "https://mcp.tracts.ai/mcp",
      "headers": {
        "x-api-key": "tracts_live_••••"
      }
    }
  }
}

Restart the client — the list_counties and get_parcels tools appear automatically.

Tools

Two tools, zero glue.

Your agent gets exactly what it needs to answer property questions — which counties are covered, and the records for any of them.

list_counties(state?: string)

Lists the US counties tracts.ai can return parcel data for, each with a configured flag. Optionally filter by 2-letter state code.

Returns

{ total, counties: [{ name, state, configured }] }
get_parcels(county: string, limit?: number)

Fetches parcel records for a county — "Name, ST" or a 5-digit FIPS code. Each record carries assessor fields (owner, situs/mailing address, values…) plus a provenance block.

Returns

{ county, state, source, sourceUrl, pulledAt, records[] }
Drop it in

Works with your client.

Point any MCP client at the endpoint. Use the x-api-key header for headless agents, or connect with OAuth — no key at all — from MCP-native clients like Claude and Cursor, which walk you through sign-in on first use.

  • Claude Desktop & Claude Code
  • Cursor (.cursor/mcp.json)
  • Any MCP-compatible host
OAuth · no api key.cursor/mcp.json
# MCP-native clients: no key — sign in on first use
{
  "mcpServers": {
    "tracts": {
      "url": "https://mcp.tracts.ai/mcp"
    }
  }
}
Why MCP

Data, not leads.
Tools, not scraping.

Give an agent a URL and it will scrape a portal badly. Give it the tracts MCP server and it queries county records the way it queries anything else — as typed tools, with the source and pull date on every answer.

Native tools, no glue code

Your agent calls list_counties and get_parcels directly — no custom HTTP wrappers, no response parsing to hand-roll. The tools appear in the palette automatically.

OAuth or API key

MCP-native clients connect with no key at all — the server returns an OAuth challenge and the client walks the user through sign-in. Headless agents pass an x-api-key header instead.

Same data, same lineage

The MCP server is a transport layer, not a different product. Every payload carries the same source, sourceUrl, and pulledAt lineage as the REST API.

County-direct, current

Records come straight from county assessors and GIS — not resold tape. Your agent gets owner, mailing, situs, valuation and parcel detail with the pull date attached.

Get started

Give your agent the data.

Get an API key and point any MCP client at the endpoint. Tell us your use case and we'll set you up.

Get an API key

Tell us your use case and we'll set you up with a key and access to the MCP server.

Data, not leads. No spam — we only reach out about your request.

Read the MCP docs