Use TransitRadar in an AI assistant
TransitRadar runs a Model Context Protocol server, so an assistant such as Claude or ChatGPT can look up live departures, follow a particular train, find where a unit is working, check a station's lifts and read how the railway did on any day since August 2026 — and cite the page it got the answer from.
The address
The server speaks MCP over streamable HTTP at one address:
https://transitradar.co.uk/mcp
It is free, read-only and needs no account. Every answer carries a link to the matching page on this site, so what the assistant tells you can be checked against the source.
Adding it to your assistant
Claude (claude.ai and the desktop app)
Open Settings → Connectors, choose Add custom connector, give it a name such as TransitRadar and paste the address above. No authentication is needed. Then ask in any conversation — “what's leaving King's Cross in the next half hour?” — and allow the tool when Claude asks.
Claude Code
One command: claude mcp add --transport http transitradar https://transitradar.co.uk/mcp
ChatGPT
Where your plan allows custom connectors, add one in ChatGPT's settings with the address above and no authentication. The option has moved between menus as the feature has developed, so if you cannot find it, OpenAI's help pages have the current route.
Cursor, VS Code and other MCP clients
Add a remote (HTTP) server with the address above. In a client configured by JSON, that
is an entry with "type": "http" and "url": "https://transitradar.co.uk/mcp".
What it can look up
Nine tools, each with an example of the kind of question that uses it:
- search_stations — turns a place into a station and its codes. “What is the station code for Manchester Piccadilly?”
- station_departures — the live departure or arrival board, with platforms, expected times and National Rail's reasons for delays. “Is the next train from Leeds to York on time?”
- find_train — searches the timetable the way the service page does: an operator and headcode, a time and a station, or two stations. “Find the 12:19 from Corby.”
- get_train — one train's calling points with booked, expected and actual times, per-stop cancellations and, today, where it was last reported. “Where is the 08:15 Edinburgh to King's Cross now?”
- find_unit — which service a unit number is working and where it last called. “What is unit 360107 doing today?”
- station_info — address, facilities, step-free access, industry codes and live lift status. “Are the lifts working at King's Cross?”
- disruptions — current incidents and planned engineering work, filterable by operator or place. “Is there any engineering work around Whitehaven?”
- daily_performance — how a completed day went: trains run, cancellations, punctuality by operator and the commonest reasons, from the daily archive. “How punctual was Northern last Friday?”
- bus_stop_departures — the next buses from any stop in Great Britain, live in London and timetabled elsewhere. Name the town as well as the stop, since many stops share a name. “When is the next bus from Trafalgar Square in London, stop X?”
What the assistant is told to be careful about
Each tool's description carries the traps this site's guides were written about, so an assistant is less likely to repeat them. A headcode is not unique: the same four characters run in several places at once. A missing forecast is not the same as on time. A lift listed as “data not available” means check before you travel, and a station with no lifts listed is not a station without lifts. A unit's position is the last station it called at, not a live fix, and the unit's allocation is a plan that can change on the day.
None of that stops an assistant getting something wrong. TransitRadar is independent and not an official source, and for anything that decides whether you catch a train, the station's screens and the operator are the source of record.
Limits and API keys
Without a key, requests are limited per connection address, at the same rate as the free
JSON API. That is plenty for one person
asking questions. If you are building something that calls it often, or you sit behind a
shared office address, get a free API key and add it to the address as https://transitradar.co.uk/mcp?api_key=YOUR_KEY, or send it as an X-Api-Key header if
your client allows headers. A key gives you your own limit.
What it deliberately does not do
- No journey planning. Multi-leg planning is the heaviest thing the site does; use the journey planner for that.
- No train formations. Coach counts and carriage lists come from a Rail Data Marketplace product whose terms do not let us pass it on, so the tools leave it out even though the site shows it. See reusing these figures.
- No route geometry or live map. A moving snapshot of the map does not help an assistant; the map is where that belongs.
- Nothing that writes. Every tool is read-only.
Where the answers come from
The tools call the same API this site is built on, so an assistant and the website always give the same answer. Live times and reasons come from National Rail's Darwin feed, movements from Network Rail, disruption notices from National Rail's KnowledgeBase, lift status from Network Rail, and buses from the Bus Open Data Service and Transport for London. Reasons are National Rail's own sentences and are marked as such. The full list, with licences, is on the data sources page. Requests are logged in the same way as any other request to the site, as described in the privacy notice.