Mapbox MCP Server
Official Mapbox MCP server: geocoding, routing, travel-time zones, and map images directly in your AI client.
- Skill Road
- Mapbox MCP Server
Categories
Description
The Mapbox MCP Server turns an AI assistant into a tool with spatial awareness. Instead of guessing addresses, distances, or travel times from memory, the assistant calls Mapbox's location services and works with real geodata. The server is maintained by Mapbox itself and implements the Model Context Protocol (MCP), so it works with Claude Code, Claude Desktop, Cursor, Codex, VS Code, or Goose.
What is the Mapbox MCP Server?
Mapbox runs a platform for maps and location services that sits inside many navigation, logistics, and real-estate applications. According to the provider, the MCP server bundles more than 30 tools behind a single interface and connects them to AI clients. That lets you answer questions such as "Which cafés are within a ten-minute walk of the central station?" or "In what order should I visit five customers most efficiently?" in plain language, without writing API calls yourself.
Features and tools
The tools fall into several groups. Search and geocoding convert addresses into coordinates and back, find places by category, and return details such as opening hours. Routing calculates directions with traffic options, travel-time matrices between several points, isochrones as reachability zones, map matching for GPS traces, and optimisation of stop order. For visualisation, the server generates static map images and, in clients that support MCP Apps, interactive maps directly in the chat.
The offline tools are especially handy. They compute purely locally and therefore cause no API calls according to the README: distance, bearing, midpoint, area, buffers, bounding boxes, line simplification, and union, intersection, and difference of polygons. For data-analysis tasks around GeoJSON this is a real advantage, because the assistant no longer has to laboriously reimplement geometric operations.
Setup
The easiest route is the hosted endpoint at https://mcp.mapbox.com/mcp. On the first connection the client walks you through an OAuth login in the browser, and no local installation is required. If you prefer to run the server yourself, install the npm package @mapbox/mcp-server or build the Docker image from the repository and provide a Mapbox access token as an environment variable. According to the documentation, the token is used locally. Guides for Claude Desktop, VS Code, Cursor, and Goose are included in the repository.
Costs and licence
The server itself is open source under the MIT licence. The API-based tools require a Mapbox account, which can be created free of charge. Which quotas are included and when paid usage begins is governed by Mapbox on its official pricing page. The offline tools work without API costs.
Who is the Mapbox MCP Server for?
It suits developers who want to explore location features with an assistant in a prototype, teams in logistics, sales, or real estate who analyse routes, catchment areas, and accessibility, and analysts who want to evaluate geodata and present it as a map. Compared with a plain web search, the server returns reproducible, structured results from a specialised mapping provider. If you already use Mapbox, you save glue code. If you do not want to use Mapbox data, you need another provider, because the server is tied to this platform.
Security and limits
Requests reach Mapbox's services, so location data and search terms leave your own machine. Only pass confidential customer addresses after reviewing Mapbox's privacy terms. Access tokens belong in environment variables rather than prompts or repositories, and a dedicated token scoped as narrowly as possible limits the damage in case of misuse. Some tools, such as place details, are still in a preview phase according to the documentation and may change.
Frequently asked questions
Is the Mapbox MCP Server free? The server is open source. Using Mapbox services depends on your account and its quotas.
Do I need my own token? With the hosted endpoint the OAuth login is enough, while local setups need a Mapbox access token.
Do interactive maps work everywhere? Only in clients that support MCP Apps. Other clients receive a static map image instead.
Requirements
Free Mapbox account, MCP client (Claude Code, Claude Desktop, Cursor, Codex, VS Code, Goose, etc.), and either an OAuth login for the hosted endpoint or a Mapbox access token for local operation (Node.js).
Installation instructions
Hosted: claude mcp add --transport http mapbox https://mcp.mapbox.com/mcp and complete the OAuth login in the browser. Local: install the npm package @mapbox/mcp-server and set MAPBOX_ACCESS_TOKEN as an environment variable.
claude mcp add --transport http mapbox https://mcp.mapbox.com/mcp
Authentication
OAuth login in the browser for the hosted endpoint; Mapbox access token as an environment variable for local operation.
Required access permissions
Permissions follow the Mapbox account or the scope of the access token. The tools read and compute location data; they do not modify customer data.
Transmitted or stored data
API-based tools send coordinates, addresses, and search terms to Mapbox. The offline geospatial tools compute locally without an API call.
Security risks
Transmission of location and address data to a third party, access tokens in insecure configurations, unexpected API costs under heavy use, and preview tools whose behaviour may change.
License and costs
- License
- MIT
- Cost
- free
The server is MIT licensed. The API-based tools use a Mapbox account with a free entry tier; quotas and paid usage are governed by the official Mapbox pricing page.
Alternatives
Not recorded yet.
At a glance
- Provider
- Mapbox
- Status
- Official server
- Deployment
- Local and remote
- Current version
- Not recorded yet.
- GitHub stars
- 357
- Last reviewed
- 30.09.2026
Repository and documentation
Categories
Supported clients
Related guides
Guides and background related to this entry.
Set up Mapbox MCP Server
Set up the Mapbox MCP Server: hosted endpoint or local token, a first test, and sensible limits.
30.09.2026
Set up the Elastic Agent Builder MCP Server
Enable Agent Builder in Kibana, configure tools, and securely connect the built-in MCP endpoint to an AI client via API key or OAuth 2.1.
20.09.2026
Set up the Searchcraft MCP Server
Start Set up the Searchcraft MCP Server with verified links, minimal permissions, and a safe first test.
19.09.2026
Set up Redis MCP Server safely
Configure Redis MCP locally, scope ACL rights, and encrypt the connection.
18.09.2026