mbta-mcp-server
A plugin that gives Claude live access to Boston-area transit data — real-time predictions, service alerts, trip planning. Built to explore how agents help you learn a new domain without becoming an expert first.
View project →The MBTA has a solid API. I had no particular need to know it deeply. But I’d been curious about a new class of AI integration — plugins that give an AI assistant live access to external data sources — and transit data makes a good domain to explore: bounded enough to understand quickly, real enough to be interesting.
So I used an agent to go learn it. I described what I wanted — real-time train predictions, service alerts, trip planning, nearby station lookup — and let the AI do the domain research while I focused on the architecture. The result is a production-quality integration: cryptographically signed releases so users can verify the software wasn’t tampered with, an automated inventory of every software component in the build, and continuous vulnerability scanning. The kind of rigor you’d apply to production software, not a weekend experiment.
That’s the point. Not mastery of the MBTA API — but that a technical leader can enter an unfamiliar domain, ship something real, and apply the same standards they’d bring to any production system. The cost of entering unfamiliar territory dropped low enough to make the exploration worth it.
This is what agentic AI changes for technical leaders. You don’t have to choose between depth and breadth anymore.