TfL
MCP server for Transport for London — lines, journeys, stop points, arrivals, bike points, occupancy, road disruptions and more over stdio
tfl-mcp
MCP server for the Transport for London Unified API — lines, journeys, stop points, arrivals, bike points, occupancy, road disruptions and more over stdio.
Installation
npx -y @daanrongen/tfl-mcp
Tools (80 total)
| Domain | Tools | Coverage |
|---|---|---|
| AccidentStats | accident_stats | Road accidents by year |
| AirQuality | air_quality | Live pollution forecasts (NO2, O3, PM10, PM2.5, SO2) |
| BikePoint | bike_points_all, bike_point_search, bike_point_by_id | Santander Cycles availability |
| Cabwise | cabwise_search | Licensed taxis & minicabs near a location |
| Journey | journey_plan, journey_modes | Full journey planner (all modes) |
| Line | 14 tools | Status, routes, disruptions, arrivals, timetables, stop sequences |
| Mode | mode_active_service_types, mode_arrivals | Cross-mode service info |
| Occupancy | 5 tools | Car parks, bike docks, EV charge connectors |
| Place | 7 tools | Search, geo lookup, postcode streets, place types |
| Road | 8 tools | TLRN status, disruptions, closures, roadworks |
| Search | 5 tools | Full-text TfL site/data search |
| StopPoint | 17 tools | Search, arrivals, disruptions, crowding, routes, taxi ranks, car parks |
| Vehicle | 3 tools | ULEZ compliance, emissions surcharge, vehicle arrival tracking |
Setup
API key (optional but recommended)
Register for a free key at https://api-portal.tfl.gov.uk/. Without one, requests are rate-limited to ~500/day.
Claude Desktop
Edit ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"tfl": {
"type": "stdio",
"command": "npx",
"args": ["-y", "@daanrongen/tfl-mcp"],
"env": {
"TFL_API_KEY": "your-key-here"
}
}
}
}
Or via the CLI:
claude mcp add tfl npx -- -y @daanrongen/tfl-mcp -e TFL_API_KEY=your-key-here
Development
bun install
bun run dev # run with --watch
bun test # run test suite
bun run build # bundle to dist/main.js
bun run inspect # open MCP Inspector in browser
Inspecting locally
bun run inspect launches the MCP Inspector against the local build:
bun run build && bun run inspect
This opens the Inspector UI in your browser where you can call any tool interactively and inspect request/response shapes.
Journey planner — location IDs
The most common failure mode is passing a free-text name to journey_plan, which causes TfL to return a 300 disambiguation response. The tool handles this gracefully and returns suggested parameterValue IDs to retry with.
Preferred ID formats (most to least reliable):
| Format | Example | Notes |
|---|---|---|
| ICS code | 1000129 | Most reliable — use output from stoppoint_search |
| Naptan ID | 940GZZLUVIC | Reliable for tube/rail stations |
| Postcode | N1C4TB | Always resolves unambiguously |
| Coordinates | 51.5308,-0.1238 | Always unambiguous |
| Free text | King's Cross | May trigger disambiguation |
Common station ICS codes:
| Station | ICS code |
|---|---|
| King's Cross St. Pancras | 1000129 |
| Victoria | 1000248 |
| Waterloo | 1000254 |
| London Bridge | 1000135 |
| Paddington | 1000184 |
| Liverpool Street | 1000134 |
| Euston | 1000078 |
| Canary Wharf | 1001006 |
| Brixton | 1000023 |
| Stratford | 1000222 |
Architecture
src/
├── config.ts # Effect Config — TFL_API_KEY
├── main.ts # Entry point — ManagedRuntime + StdioServerTransport
├── domain/
│ ├── TflClient.ts # Context.Tag service interface
│ └── errors.ts # TflError, TflDisambiguationError
├── infra/
│ ├── TflClientLive.ts # Layer.effect — HTTP client with disambiguation handling
│ └── TflClientTest.ts # In-memory test adapter
└── mcp/
├── server.ts # McpServer wired to ManagedRuntime
├── utils.ts # formatSuccess, formatError, formatDisambiguation
└── tools/ # One module per TfL domain (13 files)
相關伺服器
Feishu/Lark OpenAPI
Connect AI agents with the Feishu/Lark platform for automation, including document processing, conversation management, and calendar scheduling.
Limitless MCP
Connect your Limitless Pendant data to Claude and other LLMs using the Limitless API.
PowerPoint MCP Server
Manipulate PowerPoint presentations using the python-pptx library.
Agentify
A multi-client AI agent monitoring and control system with automatic task completion detection.
Reclaim AI
Reclaim is an ai calendar management tool that helps you plan your schedule and tasks.
Bookstack MCP
An MCP server for interacting with Bookstack, built with the mcp-framework for Node.js.
Agile Luminary
Connects AI clients to the Agile Luminary project management system via its REST API.
Umami MCP Server
Integrate Umami Analytics with any MCP client like Claude Desktop, VS Code, and more.
ResumeTailor
Automatically tailors resumes for specific job applications using LibreOffice.
context-distill
context-distill is an MCP server that compresses noisy command output into precise, actionable summaries for LLM workflows. Use distill_batch for large logs and distill_watch for cycle-to-cycle deltas. Built with Go, Cobra, Viper, and DI for reliable local and provider-backed distillation.