Goose Setup

View on GitHub

This guide explains how to configure Goose AI agent framework with the Mapbox MCP Server.

Requirements

Setup Instructions

1. Install Goose

Follow the official Goose installation instructions.

2. Configure Mapbox MCP Server

Add the Mapbox MCP Server to Goose's configuration file (typically ~/.config/goose/profiles.yaml):

default:
  provider: openai
  processor: gpt-4
  accelerator: gpt-4o-mini
  moderator: passive
  toolkits:
    - name: developer
    - name: mcp
      requires:
        servers:
          - name: mapbox
            command: npx
            args:
              - '-y'
              - '@mapbox/mcp-server'
            env:
              MAPBOX_ACCESS_TOKEN: 'your_token_here'

Alternatively, if you've built the server locally:

default:
  provider: openai
  processor: gpt-4
  accelerator: gpt-4o-mini
  moderator: passive
  toolkits:
    - name: developer
    - name: mcp
      requires:
        servers:
          - name: mapbox
            command: node
            args:
              - '/path/to/mcp-server/dist/esm/index.js'
            env:
              MAPBOX_ACCESS_TOKEN: 'your_token_here'

3. Start Goose

goose session start

4. Try It Out

Ask Goose to create a map:

Show me a map of the Golden Gate Bridge

MCP-UI Support ✨

Goose supports MCP-UI, which means you'll see inline interactive maps rendered directly in the chat when you use the static_map_image_tool. This provides a richer visual experience compared to clients that only display static images.

For more information about MCP-UI features, see the MCP-UI documentation.

Troubleshooting

Maps not appearing?

Verify that:

  1. Your Mapbox access token is correct
  2. Goose is properly configured (check ~/.config/goose/profiles.yaml)
  3. The MCP server is accessible (check Goose logs)

Want to disable MCP-UI?

Add the --disable-mcp-ui flag to args:

args:
  - '-y'
  - '@mapbox/mcp-server'
  - '--disable-mcp-ui'

For more information about Goose, visit the official Goose repository.