Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Description from the maker (original English)
Read-only Transport for London data: line status, arrivals, journeys, stop lookup.
These three are never merged into one score: opinions, measurements and popularity are different things.
We have not written a plain-language summary for this one yet. Below are the facts we read from the registry.
Pick the AI app you use, then copy the settings.
TFL_APP_KEY, TFL_MCP_MIN_INTERVAL_MSValues marked secret are keys to your account. Never share them or paste them into a chat.
You need
Which app do you use?
Pick your app above. We remember it, so next time the button says “Install in …”.
Open Claude Desktop → Settings → Developer → Edit Config, and paste this into claude_desktop_config.json
{
"mcpServers": {
"tfl-ts": {
"command": "npx",
"args": [
"-y",
"tfl-ts",
"mcp"
],
"env": {
"TFL_APP_KEY": "your-value",
"TFL_MCP_MIN_INTERVAL_MS": "your-value"
}
}
}
}Run this in your terminal
claude mcp add tfl-ts --env TFL_APP_KEY=your-value --env TFL_MCP_MIN_INTERVAL_MS=your-value -- npx -y tfl-ts mcpFill in the empty values before running it.
Paste into .cursor/mcp.json in your project, or ~/.cursor/mcp.json
{
"mcpServers": {
"tfl-ts": {
"command": "npx",
"args": [
"-y",
"tfl-ts",
"mcp"
],
"env": {
"TFL_APP_KEY": "your-value",
"TFL_MCP_MIN_INTERVAL_MS": "your-value"
}
}
}
}Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
Paste into .vscode/mcp.json in your project
{
"servers": {
"tfl-ts": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"tfl-ts",
"mcp"
],
"env": {
"TFL_APP_KEY": "your-value",
"TFL_MCP_MIN_INTERVAL_MS": "your-value"
}
}
}
}Your browser asks to open the app, then the app shows the settings and asks you to confirm. Key values are left empty for you to fill in.
This one runs on your computer, and ChatGPT can only connect to servers on the internet. Pick another app.
Your choice stays on this device in a cookie; no account needed.
TFL_MCP_MIN_INTERVAL_MS envThese settings are generated from the maker’s registry entry. If the README says otherwise, follow the README.
Runs on your computer; the tool list appears in your AI app after you install it.
We check information anyone can see; we have not read the code. Checks that do not apply are left out of the count. · Checked Nov 8, 2026
The registry name was verified with the GitHub account ghcpuman902, which owns the repo
We have not read this repo from GitHub yet
We have not read this repo from GitHub yet
This one runs on your computer; there is no internet endpoint
Runs on your computer and uses local settings instead
Runs on your computer; we can only read tool lists from internet endpoints
Compared with every active server in this category in the registry. Stars count only when the publisher owns the repo.
It passed 1 of 1 applicable basic checks. We have not read the code. Before adding a key or token, see what access the server asks for and use a key with only the permissions you need.
You need to fill in: TFL_APP_KEY
It runs on your computer. Your AI app starts it when needed, and data does not pass through the maker’s servers (unless it calls an online service itself).
The GitHub repo last changed –, and the latest version (2.15.2) reached the registry 23 days ago.
Apps that support MCP, such as Claude Desktop, Claude Code, Cursor and VS Code. You also need Node.js on your computer.
In the same category: Spec Driven Development, Remote Desktop Commander, Tethered Trust
Servers in the same category from other makers, sorted by stars.
Reading a card:★ GitHub starsiThree pillars: reviews · checks · usageiUpdated = latest version published to the registryi
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Remote Desktop Commander: use files, terminal and processes on your own computers from any AI app.
Fit, compatibility and recall checks for AI agents before they buy: verdicts, evidence, receipts.
Copied from the maker’s repo as written. PopNx MCP has not reviewed or endorsed it. HTML, images and scripts are removed. Links lead off this site.
Typed TfL client: friendly wrappers, 84 raw REST endpoints, and offline station topology for 20 rail lines.
Built with tfl-ts → Board
| tfl-ts | tfl-unified-api | tfl-journey-planner | tfl-mcp | mcp-tfl | |
|---|---|---|---|---|---|
| TypeScript types | Yes | No | No | Source only | Yes (src/index.ts) |
| Zero dependencies | Yes | No | No | No | No |
| Raw endpoint coverage | 84 | 10 of 84 | 1 of 84 | 59 of 84 | 8 of 84 |
| Offline station data | Yes | No | No | No | No |
| MCP support | Yes | No | No | Yes | Yes |
Counts use the 84 operations in the committed OpenAPI snapshot (ENDPOINT_COUNT in src/generated/endpoints.ts, pathCount in src/generated/openapi/spec.meta.json). Checked against each repo's source and the package.json published to npm on 16 October 2026.
tfl-ts publishes .d.ts files. Runtime dependencies: none. typescript is a peer dependency, for people compiling against the types. Station order and hubs ship in tfl-ts/meta, with no network call. tfl mcp is the local MCP server.
tfl-unified-api 2.0.2 is JavaScript, with no types field. It depends on moment, request, and request-promise-native. The Line module calls ten of the 84 paths: meta modes, severity, disruption categories, service types, route sequence, route by id, lines by mode, status by id, status over a date range, and status by mode. getRoutesByServiceType builds /Line/Route/{types}, which sits outside those 84 paths. db/lineColors.json is line colours, not stations.
tfl-journey-planner 1.1.2 is LiveScript compiled to JavaScript, with no types field. It depends on request and querystring. The module calls GET /Journey/JourneyResults/{from}/to/{to} only.
@daanrongen/tfl-mcp 1.4.3 is written in TypeScript. The published tarball is dist/main.js, package.json, and README.md, and package.json has no types field. It depends on @modelcontextprotocol/sdk, effect, and zod. Its tools call 59 of the 84 paths. They also call /Road/all/Disruption (the snapshot path is /Road/all/Disruption/{disruptionIds}), /Vehicle/EmissionSurcharge, and /Vehicle/UlezCompliance.
@pipeworx/mcp-tfl 0.1.2 sets "types": "src/index.ts", and that file is in the tarball. It depends on @modelcontextprotocol/sdk and tsx. Six tools call 8 of the 84 paths.
Station order ships with tfl-ts, in tfl-ts/meta.
Get a Primary key from the TfL API Portal. Subscribe to "500 Requests per min", then Profile → Show.
pnpm add tfl-ts
export TFL_APP_KEY=your-primary-key pnpm exec tfl raw line.statusByIds --ids victoria
[
{ "id": "victoria", "name": "Victoria", "lineStatuses": [{ "statusSeverityDescription": "Good Service" }] } ]
Line names, station order, and colours ship in the package. Status, arrivals, and journeys hit TfL at runtime.
On the portal you'll see two keys (Primary and Secondary); either works. app_id has been unused since Jan 2021.
Pass credentials in the constructor instead
const client = new TflClient({
appKey: 'your-primary-key', });
import TflClient from 'tfl-ts';
const client = new TflClient(); // reads TFL_APP_KEY from process.env
const status = await client.line.getStatus({ modes: ['tube'] });
const arrivals = await client.stopPoint.getArrivals({ stopPointIds: ['940GZZLUOXC'], // Oxford Circus });
const journey = await client.journey.plan({ from: '940GZZLUOXC', to: '940GZZLUBND', });
const { matches } = await client.stopPoint.search({ query: 'Oxford Circus', modes: ['tube'], });
const busStops = await client.stopPoint.searchBusStops('Trafalgar Sq');
Server Component with ISR (~60s). Boards and explorer: tfl-components · tfl.manglekuo.com/docs/explorer. Tube boards use official line colours; bus boards use route-number chips. Do not mix. See examples/README.md.
import TflClient, { sortLinesBySeverityAndOrder } from 'tfl-ts';
export const revalidate = 60;
const client = new TflClient(); const tube = await client.line.getStatus({ modes: ['tube'] }); const sorted = sortLinesBySeverityAndOrder(tube);
pnpm dlx shadcn@latest add https://tfl.manglekuo.com/r/tube-status-board.json
pnpm dlx shadcn@latest add https://tfl.manglekuo.com/r/tfl-roundel.json
The roundel ships a placeholder unless NEXT_PUBLIC_ALLOW_TFL_ROUNDEL=true (you accept trademark responsibility).
Read-only MCP (npx tfl-ts mcp), cached and rate-limited. Static tools never call TfL. The npm bin is tfl (dist/cjs/bin/tfl.js); mcp is the stdio subcommand. Setup: docs/mcp.md.
{
"mcpServers": { "tfl-ts": { "command": "npx", "args": ["-y", "tfl-ts@latest", "mcp"], "env": { "TFL_APP_KEY": "your-primary-key" } } } }
Tools: get_supported_modes, resolve_line_id, docs, resolve_stop_id, get_line_status, get_arrivals, plan_journey.
No repo access, or lost your context? npx tfl-ts docs ls lists every bundled agent doc (this README, CLAUDE.md, docs/agent.md, docs/mcp.md, …). docs cat prints one, and docs find / docs grep search across all of them — offline, no API key. The local MCP docs tool reads the same catalogue.
'central', 'victoria', 'elizabeth'. Not display names like 'Central'.'940GZZLUOXC'. Resolve with stopPoint.search() or place.search().'51800').LINE_NAMES, STATION_SEQUENCES, mode lists) before live calls.accidentStats and airQuality are deprecated.STATION_HUBS tracks Southeastern, South Western Railway, and similar operators for topology, but getArrivals() returns an empty array for them, not an error.lineStatuses. Do not read [0]. Use getWorstCurrentStatus(line.lineStatuses) for the operative row.validityPeriods[].isNow follows disruption.category === 'RealTime'. It is not a clock check. Planned engineering can be in force today with isNow: false.validityPeriods[].toDate is the window end, not when trains resume. Weekend engineering often ends at 00:29Z (01:29 London, end of the traffic day). Overnight-split slices on one row are one possession; do not treat the first overlapping toDate as the next train.sortLinesBySeverityAndOrder ranks it after incidents.lineId flips along shared track. Use withSharedTrackIdentity. Do not rewrite raw lineId.getDetailedStatus() returns shorter types and renamed fields (severity, from, closureType, originName) instead of TfL's generated shapes:
const lines = await client.line.getDetailedStatus({
lineIds: ['bakerloo'], dateRange: { startDate: '2026-08-08', endDate: '2026-08-10' }, }); const s = lines[0]?.statuses?.[0]; s?.severity; s?.severityDescription; s?.validityPeriods?.[0]?.from; s?.disruption?.closureType;
Use getStatus({ detail: true }) or client.raw.line.* for exact TfL field names.
Tube, Elizabeth line, DLR, Overground, Tram, and river-bus pier topology. No credentials, no network. Identity, order, and branches only (no status or arrivals). River piers are not in STATION_HUBS — poll the NaptanFerryPort id. Live topology: client.line.getRouteSequence(). Also on the client: client.line.STATION_SEQUENCES.
import { LINE_STATION_SEQUENCES } from 'tfl-ts/meta';
const bakerloo = LINE_STATION_SEQUENCES.bakerloo; const outbound = bakerloo.orderedRoutes.find( (route) => route.direction === 'outbound' && route.serviceType === 'Regular', ); console.log(outbound?.stationIds, bakerloo.branches);
STATION_HUBS maps each physical station to its sibling StopPoint ids and the specific id that carries arrivals for each line — Liverpool Street's tube id (940GZZLULVT) and rail id (910GLIVST) both resolve to one HUBLST entry, with Central on the tube id and Elizabeth line on the rail id. No credentials, no network.
import { STATION_HUBS, resolveArrivalsStopId } from 'tfl-ts/meta';
const hub = STATION_HUBS['940GZZLULVT']; // any sibling id works const elizabethStopId = hub && resolveArrivalsStopId(hub, 'elizabeth'); // '910GLIVST'
resolveArrivalsStopId returns undefined when the hub doesn't carry that line, rather than the interchange id — polling the interchange id itself returns zero arrivals from TfL. Third-party National Rail operators (Southeastern, South Western Railway, c2c, and similar) show up in the hub's topology but never return live predictions: TfL's Arrivals API only covers tube, DLR, tram, Overground, and Elizabeth line.
client.stopPoint.getNormalizedArrivals() is getArrivals() plus a cleaned destination (falls through empty or literal "null" towards to destinationName, common on Elizabeth line, Overground, and some bus termini) and platform (compass bound, cleaned label, isUnknown for TfL's literal "Platform Unknown"):
const arrivals = await client.stopPoint.getNormalizedArrivals({
stopPointIds: ['940GZZLUOXC'], }); arrivals[0]?.destination.name; arrivals[0]?.platform.label;
getArrivals() and client.raw.* are unchanged.
client.stopPoint.get and getByGeoPoint lift Direction additional properties onto the stop: towards, compassPoint, and compassBearingDegrees. smsCode is filled from the first-class field or bag SmsCode. Facility keys stay in additionalProperties as TfL strings. parseAdditionalPropertyValue turns a bag value into null / boolean / number / date / text ("yes", "null", unix ms InstallDate). client.raw.stopPoint.* is unlifted. Do not use Prediction bearing (vehicle heading) as the stop flag; a painted `stopLette
Only the beginning is shown; read the rest in the repo