Build an agent with Mapbox MCP
summary
Mapbox releases a Model Context Protocol server that lets agents call Mapbox APIs directly; a demo wires Mapbox MCP Server into a Crew AI flow using Crew AI + GPT-4.1, Flask and GL JS, requiring only a Mapbox access token. The build creates three agents and three tasks that produce GeoJSON output and a camera command sequence for map animation; POI search and directions tools work and render on the map, static‑image tool misfires occasionally, and simplified route geometries can be imprecise. Code lands on Mapbox’s GitHub.
AI-written summary · about
chapters (9)
- 0:00Introduction and logistics
- 1:08Mapbox MCP overview
- 2:23Mapbox MCP use cases
- 6:05Building a basic agent
- 12:08Adding a visual interface
- 15:17Adding GeoJSON functionality
- 20:27Choreographing camera actions
- 28:28Refining and scaling the agent
- 37:18Q&A and closing remarks
related talks
23mMCP Mashups: How AI Agents are Reviving the Programmable Web
40mBuilding Agents with Agent Bricks and MCP
29mModern Day Mashups: AI Agents are Reviving the Programmable Web
39mBuilding Agentic AI Apps with Docker MCP & Cagent
13mMapStory: Prototyping Editable Map Animations with LLM Agents
28mAgentic AI systems using Model Context Protocol
21mBuilding Agents (the hard parts!)
01h 14mBuilding Next Gen Agents with MCP