MCP server

Territas speaks MCP, so an agent can underwrite a property as a tool call and reason over the cited result. It exposes two tools, mirroring the REST endpoints one-for-one:

• underwrite_parcel — one property. Same arguments as POST /v1/underwrite: address, purchase_price, annual_rent, annual_taxes, annual_insurance_base, plus optional loan_amount, structure_value, residential, and as_of.
• bulk_underwrite_parcels — a list of up to 500 properties with an optional concurrency, returning the same summary + results shape as bulk underwrite.

Both return the full survey — flood determination, risk-adjusted NOI, and the source audit trail — so the model’s answer stays grounded in federal data rather than a guess.

Connecting a client

Once the remote server is live, you’ll point any MCP client at it and authenticate with the same X-API-Key you use for REST — the key carries your tier and metered usage, so MCP calls bill exactly like API calls. A client config will look like this:

{
  "mcpServers": {
    "territas": {
      "url": "https://mcp.territas.com",
      "headers": {
        "X-API-Key": "trs_…your-key…"
      }
    }
  }
}

Then ask in plain language, and the agent picks the tool:

Underwrite 1600 Pennsylvania Ave NW, Washington DC 20500.
Purchase price 450k, annual rent 42k, taxes 6.2k, base insurance 1.8k.
Is NFIP flood insurance required, and what's the risk-adjusted NOI?

Metering and limits

MCP tool calls are underwrites: underwrite_parcel meters one, and bulk_underwrite_parcels meters one per successful property, exactly as the REST endpoints do. Your tier’s rate limit and included volume apply the same way — see Plans & limits.