BVG MCP Server – Unofficial

The BVG MCP Server fills a niche for developers who want to give their AI assistants real‑time access to public transport data in Berlin. By exposing the BVG API through the Model Context Protocol, the server allows models such as Claude to query bus and tram stops, schedules, and other transit information without embedding the API logic directly into the assistant. This abstraction reduces maintenance overhead for developers, keeps sensitive credentials out of the model’s prompt space, and ensures that the assistant can be updated simply by changing the server rather than retraining or re‑prompting.

At its core, the server implements a single STDIO transport variant of MCP. Each time a client like Claude Desktop initiates a session, the server spawns a fresh process that listens on standard input and writes responses to standard output. This design keeps the server lightweight, eliminates network configuration for local development, and guarantees that a client’s lifecycle is tightly coupled to the server process – it exits automatically when the user closes the assistant. The simplicity of STDIO makes it compatible with any MCP‑compliant client that can spawn a subprocess, which is especially useful for desktop or command‑line tools.

The server currently offers one tool: Stop Information. When invoked, it accepts a location string (e.g., “Turmstrasse”) and returns the name of the corresponding BVG stop. While minimal, this tool demonstrates how an MCP server can translate natural language queries into structured API calls and return clean, machine‑readable data. Developers can extend the server to include additional endpoints such as timetable lookups, real‑time arrival predictions, or route planning, all while keeping the MCP interface unchanged.

Real‑world use cases are plentiful. A travel assistant could let users ask, “What’s the next tram at Turmstrasse?” and receive a precise arrival time without exposing the BVG credentials to the model. A logistics platform might integrate the server into its AI‑driven dispatch system, allowing planners to query public transport availability on demand. Because the server runs locally, it respects privacy constraints and can be deployed behind corporate firewalls without exposing external endpoints.

In summary, the BVG MCP Server is a lightweight bridge between an AI assistant and Berlin’s public transport data. Its STDIO‑based design, straightforward tool interface, and ease of integration make it an attractive choice for developers who need reliable, secure access to BVG information within their MCP workflows.