🗺️ OpenRoute MCP

Model Context Protocol (MCP) server to plan routes using the OpenRouteService API, for activities such as hiking or mountain biking.

🛠️ Available tools

• 📍 Find possible coordinates for a location

  • Arguments:
    • location (string): location to get coordinates for
  • Returns: list of 10 coordinates with available description to enable the agent figure out the right one

• 🏠 Find possible locations (addresses) for given coordinates.

  • Arguments:

    • lon (float): Longitude of the location
    • lat (float): Latitude of the location

  • Returns: list of objects close to the given coordinates

• 🗺️ Create a route from a starting location to a destination, optionally with waypoints

  • Arguments:
    • route_type (string): Type of route, e.g. "cycling-mountain", "cycling-regular", "foot-hiking", "driving-car"
    • from_coordinates (list[float]): Starting location as [longitude, latitude]
    • from_coordinates (list[float]): Destination location as [longitude, latitude]
    • waypoints (list[list[float]]): optional list of waypoints coordinates as [[lon, lat], ...]
  • Returns:
    • a resource file containing the route GPX (e.g. route://foot-hiking-98.gpx)
    • a resource file containing a PNG image visualization of the route (e.g. route://foot-hiking-98.png)
    • a resource file containing a HTML interactive visualization of the route (e.g. route://foot-hiking-98.html) that can be directly opened with your browser

• 📌 Search for Points of Interest (POIs) in an area

  • Arguments:
    • bounding_box_coordinates (list[list[float]]): coordinates defining a bounding box as [[min_lon, min_lat], [max_lon, max_lat]]
    • filters_name (list[string]): optional list of names to filter POIs, e.g. ["Gas station", "Restaurant"]
  • Returns: found POIs information

• 🔎 Search for known routes close to a given start and end coordinates, this enables to suggest waypoints to go through known trails when creating a route. It is only working in Switzerland for now thanks to the Swiss Confederation api3.geo.admin.ch 🇨🇭

  • Arguments:
    • route_type (string): Type of route, e.g. "cycling-mountain", "cycling-regular", "foot-hiking", "driving-car"
    • from_coordinates (list[float]): Starting location as [longitude, latitude]
    • from_coordinates (list[float]): Destination location as [longitude, latitude]
  • Returns: known trails trace and description when available to help the agent build a nice route

• ⏱️ Computes the area that can be reached within a given time or distance from one or more starting points.

  • Arguments:

    • coordinates_list: 1 or more coordinates to compute reachable area from as [[lon, lat], ...]
    • route_type: Type of route, e.g. "cycling-mountain", "cycling-regular", "foot-hiking", "driving-car"
    • range_type: Type of range, either time (in seconds) or distance (in metres)
    • area_range: maximum range value of the analysis in seconds for time and metres for distance. Or a comma separated list of specific range values

  • Returns: Reachable area information in GeoJSON format

🗃️ Available Resources

• 🚏 Route files
  • URL pattern: route://{filename}
  • Type: GPX (.gpx), HTML (.html), or PNG (.png)
  • Description: Returns a previously generated route file by filename.
  • Produced by: create_route_from_to tool

🔌 Connect client to MCP server

Follow the instructions of your client, and use the /mcp URL of your deployed server (e.g. http://localhost:8888/mcp)

For example, for GitHub Copilot in VSCode, to add a new MCP server through the VSCode UI:

• Open side panel chat (ctrl+shift+i or cmd+shift+i), and make sure the mode is set to Agent in the bottom right
• Open command palette (ctrl+shift+p or cmd+shift+p), and search for MCP: Open User Configuration, this will open a mcp.json file

⌨️ Use stdio transport

OPENROUTESERVICE_API_KEY=YYY uvx openroute-mcp

In VSCode mcp.json file you should have:

{
  "servers": {
    "openroute-mcp": {
      "command": "uvx",
      "args": [
        "openroute-mcp"
      ],
      "env": {
        "OPENROUTESERVICE_API_KEY": "YOUR_API_KEY"
      }
    }
  }
}

Full details of the CLI options:

usage: openroute-mcp [-h] [--http] [--port PORT] [--openroute-api OPENROUTE_API] [--openroute-api-key OPENROUTE_API_KEY]
                     [--data-folder DATA_FOLDER] [--no-save] [--no-img] [--add-html]

A Model Context Protocol (MCP) server for building routes using OpenRouteService.

options:
  -h, --help            show this help message and exit
  --http                Use Streamable HTTP transport
  --port PORT           Port to run the server on
  --openroute-api OPENROUTE_API
                        OpenRouteService API URL (default: https://api.openrouteservice.org)
  --openroute-api-key OPENROUTE_API_KEY
                        OpenRouteService API key (default: taken from env var OPENROUTESERVICE_API_KEY)
  --data-folder DATA_FOLDER
                        Folder to save generated routes
  --no-save             Don't save generated routes to disk (also disable image and HTML generation)
  --no-img              Do not add PNG image visualization of the routes to the response (image not supported by all LLMs)
  --add-html            Add HTML interactive map for routes to the response (larger context used)

📡 Use streamable HTTP server

Connect to a running streamable HTTP MCP server, such as openroute-mcp.onrender.com/mcp

In VSCode mcp.json you should have the following:

{
  "servers": {
    "openroute-mcp-http": {
      "url": "https://openroute-mcp.onrender.com/mcp",
      "type": "http"
    }
  }
}

More details available in the VSCode MCP official docs.

🧑‍💻 Development

Checkout the CONTRIBUTING.md page for more details on running in development and contributing.