Mapricot MCP server

Mapricot's MCP server lets AI agents and coding assistants geocode addresses, find places, plan routes, locate IP addresses, and create browser keys for your site. It's a remote server, so there's nothing to install: add it to your agent, then sign in to Mapricot, with no API key to copy.

Connect your agent

Claude

1. In Claude, open Settings, then Connectors, and choose Add custom connector.
2. Name it Mapricot, enter https://api.mapricot.com/mcp as the server URL, and press Add.
3. Press Connect next to Mapricot, and sign in to Mapricot in the window that opens.

On a Team or Enterprise plan, an owner adds the connector first, in Organization settings, then Connectors. Then each member connects to it from their own settings.

Claude Code

Add the server from your terminal:

claude mcp add --transport http mapricot https://api.mapricot.com/mcp

Then, in Claude Code, run /mcp, pick Mapricot, choose Authenticate, and sign in to Mapricot in the browser window that opens.

Add --scope user to the command to use Mapricot in every project, not just the current one.

ChatGPT

1. Open Plugins in ChatGPT, press the + button, and choose Add custom MCP server.
2. Name it Mapricot, and under Connection, enter https://api.mapricot.com/mcp.
3. Choose OAuth for authentication, read the warning, and choose I understand and want to continue, then Create as a plugin.
4. Install the plugin, and sign in to Mapricot in the window that opens.
5. In a new chat, type @ and pick Mapricot.

On a Business or Enterprise workspace, an admin may need to allow custom MCP servers first.

Cursor

Add Mapricot to Cursor in one click, or add it yourself to ~/.cursor/mcp.json, for every project, or to .cursor/mcp.json in one project:

{
  "mcpServers": {
    "mapricot": {
      "url": "https://api.mapricot.com/mcp"
    }
  }
}

Cursor asks you to sign in to Mapricot the first time it uses the server.

VS Code

Add it to .vscode/mcp.json in your project, or run MCP: Add Server from the Command Palette, choose HTTP, and paste https://api.mapricot.com/mcp:

{
  "servers": {
    "mapricot": {
      "type": "http",
      "url": "https://api.mapricot.com/mcp"
    }
  }
}

Codex

Add the server, then sign in. The second command opens Mapricot in your browser:

codex mcp add mapricot --url https://api.mapricot.com/mcp
codex mcp login mapricot

Other clients

Any client that connects to remote MCP servers over Streamable HTTP and signs in with OAuth works: give it https://api.mapricot.com/mcp. The server speaks MCP 2025-06-18 and later. Clients find where to sign in from the server's 401, and register with a Client ID Metadata Document or, if they don't publish one, through dynamic client registration.

Signing in

The first time your agent uses Mapricot, a browser window opens to sign you in. If you're new to Mapricot, you can sign up there, with the same 250,000 free requests every 30 days.

The sign-in page shows which app is asking. For an app that publishes its own metadata, such as Claude, it names the domain the app comes from, which the app can't fake. An app that registers itself can call itself anything, so the page says Mapricot can't check its name and shows the address you go back to. Only connect an app you just started connecting.

You pick:

• An account. Every request the agent makes is billed to it, like any API key's.
• What it can do: the products the agent can use, and whether it can create browser keys. Creating keys starts unticked.

Allowing it adds an agent key to the account's API keys page, named after the app, with the app's domain beside the name. Its usage shows with the account's other keys. The agent never sees a key: it gets a token that lasts an hour and renews itself while the agent key is live. If the agent goes 30 days without using Mapricot, it asks you to sign in again. If you leave the account, the agent loses access too.

Tools

The agent sees only the tools for the products you allowed. Each tool sends one API request with the agent's arguments and gives back the API's JSON, with its attribution.

There are no tools for autocomplete or batches, since agents send whole queries one at a time. A map the agent builds into your site loads Map tiles with a browser key.

Maps in the conversation

If you allowed Map tiles, apps that support MCP Apps show maps in the conversation, which you can pan and zoom. Routes and reachable areas come with their own map. For anything else, such as "show me coffee shops in Portland", the agent looks the places up and then calls show_map to put them all on one map; click a place to see its name and details. show_map isn't billed. In other apps, the agent gives you the results as text.

Billing and limits

Calls to search, reverse, nearby, place, route, optimize_route, isochrone, and ip_location are API requests like any other: they count toward your free requests and credit, at the same price, and toward the agent key's caps and your account's limits.

To limit what an agent can spend, edit its agent key on the API keys page and set its requests per minute or per 30 days. A request over a cap, or with no credit left, comes back to the agent as an error it can read, such as period_cap_reached or credit_exhausted, the same as the API's errors.

Browser keys from agents

If you let it create browser keys, an agent can finish a job like "add a map to my site" on its own. create_key makes a browser key in the agent's account, for the origins the agent names, such as https://shop.example.com and http://localhost:*, with products from among the agent's own. The agent gets the key once, to put in your site's code.

• Never a server key. A server key's secret would end up in the chat. If your backend needs one, the agent sends you to the API keys page to create it and copy it into your .env yourself.
• No more than the agent's caps. If you capped the agent key's requests per minute or per 30 days, its keys get those caps unless it sets lower ones. Otherwise they have only the caps the agent sets. Its tool description suggests a cap per IP address per hour, since a browser key is public in your page.
• Ten an hour. Agents can create ten keys an hour in an account.

The API keys page shows each key an agent made as "Browser key made by" that agent, and the audit log records which app created it. Revoking the agent key leaves the keys it made working, since your site may already use them, so revoke those yourself if you don't want them.

Disconnect an agent

Revoke its agent key on the API keys page. Its tokens stop working on its next request. To connect it again, sign in from the agent again, which makes a new agent key.

An app that registered itself, rather than publishing its own metadata, has its registration deleted after 30 days with no agent key and no sign-in. If the sign-in page says Mapricot no longer has the app's registration, remove Mapricot from the app and add it again.