An MCP server for use with LLM chatbots providing tools related to OpenStreetMap
This MCP server can be integrated with your client that supports MCP.
The docker-compose.yaml
file included will run both the púca UI and MCP server so that you can enter queries into the UI.
To do so you will need to add your own OpenAI API Key in the .env
file then start the containers with:
docker compose up -d
Check the logs for any issues and visit the web interface at http://your.ip.address:3301
See the example file client.py
for how to use this MCP server with Python's pydantic_ai
Anthropic's Claude supports adding MCP servers if you edit the file:
vi ~/Library/Application\ Support/Claude/claude_desktop_config.json
adding the following but amending the URL:
{
"mcpServers": {
"puca": {
"command": "npx",
"args": [
"mcp-remote",
"http://192.168.1.40:3300/sse"
]
}
}
}
Below is a list of the tools made available via the MCP server:
- get_coordinates_for_address
- get_address_by_coordinates
- get_defibrillators
- get_distance_between_addresses
- get_distance_between_coords
- get_parking
- get_toilets
- get_post_offices
- get_cafes
- get_fast_food_places
- get_irish_street_names
- get_vacant_buildings
- get_overpass_results
This MCP server makes use of the freely available APIs from OSRM, Overpass and Nominatum. The data in OpenStreetMap is contributed by dedicated volunteers.