AI agents

Give an AI agent real knowledge of Nigeria's streets, places and roads. MilsatMaps ships an MCP server, so Claude, ChatGPT, Cursor or any client that speaks the Model Context Protocol can search places, read them, say what's at a point and plan a route, with answers grounded in the same map your app shows.

The tools

ToolWhat the agent can do
search_placesFind addresses, streets, landmarks, businesses and postcodes, optionally near a point.
place_detailsRead one place in full: category, area, opening hours and other attributes, contacts, photos.
reverse_geocodeSay what's at a point: a one-line address, the street, the area and the postcode.
get_directionsPlan a driving or walking route: distance, time and turn-by-turn steps.

Set up

Make a secret key in the console (the agent runs on your machine or server, never in a web page), then add the server to your client's MCP settings. In Claude Desktop, Claude Code and Cursor it looks like this:

{
  "mcpServers": {
    "milsatmaps": {
      "command": "npx",
      "args": ["-y", "@milsat/mcp"],
      "env": { "MILSAT_API_KEY": "sk_live_…" }
    }
  }
}

Then ask in plain words. The agent picks the tools itself:

Find a pharmacy near Yaba market and tell me how long it takes to drive
there from Ikeja City Mall.

In your own agent

Building your own agent with a model API? Run the same server as a subprocess and pass its tools to the model, or call the REST API from tools you define: the four tools map one to one onto /v1/search, /v1/places/{id}, /v1/reverse and /v1/route.

Usage and credit

Each tool call is one request against your project's allowance, like any API call, and shows on the console's Usage page. Where your agent shows results to people, credit them as Attribution and data describes.