Spec Driven Development
MCP Server that facilitates spec-driven development workflows, not just Vibe Coding.
Beschreibung vom Entwickler (englisches Original)
Static FE-BE contract-drift checker: finds frontend calls the backend does not expose. CLI + MCP.
Diese drei werden nie zu einer Punktzahl zusammengefasst: Meinungen, Messungen und Beliebtheit sind verschiedene Dinge.
Für diesen Server haben wir noch keine Zusammenfassung in einfacher Sprache geschrieben. Unten stehen die Fakten, die wir aus der Registry gelesen haben.
Wähle die KI-App, die du nutzt, und kopiere die Einstellungen.
Du brauchst
Welche App nutzt du?
Wähle oben deine App. Wir merken sie uns, damit der Button beim nächsten Mal „In … installieren“ anzeigt.
Öffne Claude Desktop → Settings → Developer → Edit Config und füge dies in claude_desktop_config.json ein
{
"mcpServers": {
"tieline": {
"command": "npx",
"args": [
"-y",
"@nugehs/tieline",
"mcp"
]
}
}
}Führe das in deinem Terminal aus
claude mcp add tieline -- npx -y @nugehs/tieline mcpIn .cursor/mcp.json in deinem Projekt oder in ~/.cursor/mcp.json einfügen
{
"mcpServers": {
"tieline": {
"command": "npx",
"args": [
"-y",
"@nugehs/tieline",
"mcp"
]
}
}
}Dein Browser fragt, ob er die App öffnen darf, dann zeigt die App die Einstellungen und bittet dich um Bestätigung. Schlüsselwerte bleiben leer, damit du sie selbst einträgst.
In .vscode/mcp.json in deinem Projekt einfügen
{
"servers": {
"tieline": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@nugehs/tieline",
"mcp"
]
}
}
}Dein Browser fragt, ob er die App öffnen darf, dann zeigt die App die Einstellungen und bittet dich um Bestätigung. Schlüsselwerte bleiben leer, damit du sie selbst einträgst.
Dieser Server läuft auf deinem Computer, und ChatGPT kann sich nur mit Servern im Internet verbinden. Wähle eine andere App.
Deine Auswahl bleibt in einem Cookie auf diesem Gerät; kein Konto nötig.
Diese Einstellungen werden aus dem Registry-Eintrag des Entwicklers erzeugt. Wenn in der README etwas anderes steht, folge der README.
Läuft auf deinem Computer; die Tool-Liste erscheint nach der Installation in deiner KI-App.
Wir prüfen Informationen, die jeder sehen kann; den Code haben wir nicht gelesen. Checks, die nicht zutreffen, werden nicht mitgezählt. · Geprüft am 7. Nov. 2026
Der Registry-Name wurde mit dem GitHub-Konto nugehs bestätigt, dem das Repo gehört
Dieses Repo haben wir noch nicht von GitHub gelesen
Dieses Repo haben wir noch nicht von GitHub gelesen
Dieser Server läuft auf deinem Computer; es gibt keinen Endpunkt im Internet
Läuft auf deinem Computer und nutzt stattdessen lokale Einstellungen
Läuft auf deinem Computer; Tool-Listen können wir nur von Endpunkten im Internet lesen
Verglichen mit allen aktiven Servern dieser Kategorie in der Registry. Sterne zählen nur, wenn dem Herausgeber das Repo gehört.
Er hat 1 von 1 zutreffenden Basischecks bestanden. Den Code haben wir nicht gelesen. Bevor du einen Schlüssel oder Token hinzufügst, sieh nach, welche Zugriffe der Server verlangt, und nutze einen Schlüssel, der nur die Berechtigungen hat, die du brauchst.
Die Registry nennt keinen erforderlichen Schlüssel, manche Server bitten dich aber bei der ersten Nutzung, dich anzumelden.
Er läuft auf deinem Computer. Deine KI-App startet ihn bei Bedarf, und die Daten laufen nicht über die Server des Entwicklers (außer der Server ruft selbst einen Onlinedienst auf).
Das GitHub-Repo wurde zuletzt – geändert, und die neueste Version (0.1.4) kam vor 5 Monaten in die Registry.
Apps, die MCP unterstützen, etwa Claude Desktop, Claude Code, Cursor und VS Code. Außerdem brauchst du Node.js auf deinem Computer.
In derselben Kategorie: Spec Driven Development, Remote Desktop Commander, First Sale Scout
Server derselben Kategorie von anderen Entwicklern, sortiert nach Sternen.
So liest du eine Karte:★ GitHub-SterneiDrei Säulen: Bewertungen · Checks · NutzungiAktualisiert = neueste Version in der Registry veröffentlichti
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.
Playbooks and a 14-day checklist for founders winning their first paying customers.
Unverändert aus dem Repo des Entwicklers übernommen. PopNx MCP hat den Inhalt weder geprüft noch empfiehlt es ihn. HTML, Bilder und Skripte wurden entfernt. Links führen von dieser Website weg.
Static frontend↔backend contract-drift checker. Pact without writing a single contract test.
tieline reads the code you already wrote on both sides of an API boundary — the HTTP calls your frontend makes and the routes your backend exposes — and tells you where they disagree. No contract tests to author, no broker to run, no backend to boot. It finishes in well under a second and is built to run in CI as a gate.
Delete the LLM and a developer still installs it.
tielineis a deterministic
tool first; an agent reading its output is a bonus.
frontend code backend code
(axios, rtk-query, …) (express, nestjs, openapi, …) │ │ client adapter server adapter │ normalize ${id} :id {id} → {} │ └───────────────► join on (METHOD, path) ◄──────┘ │ ✅ matched ❌ drift ⚠️ unverifiable 🟡 dead
Both sides are reduced to a canonical (METHOD, path) and joined. The matcher is adapter-agnostic — any client adapter pairs with any server adapter, so the same engine checks a React app against Express, an Angular app against Spring, or anything against an OpenAPI spec.
| Result | Meaning |
|---|---|
| ✅ matched | FE call resolves to a real BE route |
| ❌ drift | FE call resolves but the BE has no such route/method — the bug bucket |
| ⚠️ unverifiable | FE url is built at runtime — reported, never guessed |
| 🟡 dead | BE route no resolvable FE call reaches (informational) |
These tools solve neighbouring problems — pick by what you have and what you want guaranteed.
| Tool | How it works | What it catches | What it needs | Reach for it when |
|---|---|---|---|---|
| tieline | Static analysis of FE call sites + BE route declarations, joined on (METHOD, path) | A frontend calling a route the backend doesn't expose (path/method drift), undocumented/phantom spec routes | Source code of both sides; nothing running, nothing authored | You want a sub-second CI gate with zero contract tests to write or maintain |
| [Pact](https://pact.io) | Consumer-driven contract tests executed at runtime, shared via a broker | Request/response payload mismatches between specific consumer–provider pairs | Contract tests written on both sides, a Pact broker, provider verification builds | Independent teams need payload-level guarantees and a can-I-deploy workflow |
| [openapi-diff](https://github.com/OpenAPITools/openapi-diff) | Diffs two OpenAPI documents | Breaking changes between two spec versions | Accurate specs for both versions; doesn't read source code | Your API surface is spec-first and you want to gate spec changes |
| [Optic](https://www.useoptic.com) | Tracks your OpenAPI spec over time and diffs every change in CI | Breaking changes and style/standards violations in the spec's history | An OpenAPI spec kept in the repo | You govern an evolving spec and want each PR's API changes reviewed |
| [Schemathesis](https://schemathesis.io) | Property-based fuzzing of a running API against its OpenAPI/GraphQL schema | Server crashes, schema violations, undocumented responses at runtime | A bootable backend + a schema | You want runtime conformance and robustness testing of the implementation |
They compose: tieline catches FE↔BE drift statically in every PR, tieline doctor keeps code↔spec honest, and a runtime tool like Pact or Schemathesis can guard the payload/behaviour layer underneath.
Install (or run with npx):
npm install -g @nugehs/tieline
Generate a tieline.config.json — init sniffs the surrounding directories (cwd, its children, and its siblings) for known stacks and writes a ready-to-run config:
tieline init
tieline · init
scanning ~/code for repos… ✔ client: rtk-query → web (roots: src/redux/apis) ✔ server: nestjs → api (roots: src)
📝 wrote tieline.config.json
It detects adapters from package.json deps (@reduxjs/toolkit, axios, @angular/core, @nestjs/core, express, fastify, next), requirements.txt / pyproject.toml (fastapi, flask), pom.xml / build.gradle (spring), and any OpenAPI doc as a fallback. Anything it can't detect is written as a placeholder for you to edit. The file is always overwritten, so re-run it whenever your layout changes.
Or write it by hand — at the root of (or above) your repos:
{
"client": { "adapter": "rtk-query", "repo": "../web", "roots": ["src/redux/apis"], "basePath": "/api/v1" }, "server": { "adapter": "nestjs", "repo": "../api", "roots": ["src"], "globalPrefix": "api/v1" }, "failOn": ["drift"] }
Run it:
tieline check
tieline · contract check
❌ 2 drift (FE calls a route the backend does not expose) GET /users/{} getUser · web/src/redux/apis/user-api.ts:42 → did you mean "user/{}"? # plural vs singular — a guaranteed 404 PUT /orders/{} updateOrder · web/src/redux/apis/order-api.ts:88 → path exists but as GET, not PUT # method mismatch
✅ 274 matched ❌ 2 drift ⚠️ 6 unverifiable 🟡 31 unused backend routes
check exits non-zero when any failOn bucket is non-empty — drop it into CI and the build fails the moment the two sides disagree.
tieline init # auto-detect nearby repos and write tieline.config.json
tieline check # FE↔BE drift; exits non-zero on drift (the CI gate) tieline list # the full resolved contract map (every endpoint + status) tieline orphans # backend routes no frontend call reaches tieline doctor # code↔spec drift (see below)
| Flag | Effect |
|---|---|
--config | Path to tieline.config.json (default: searched upward from cwd) |
--json | Machine-readable output |
--html | Self-contained visual report (see Visual report) |
--no-fail | Always exit 0 (report only) |
Any client adapter pairs with any server adapter — the matcher never changes.
| Client (calls) | Server (routes) |
|---|---|
rtk-query — Redux Toolkit Query | nestjs — decorators |
axios-fetch — axios / fetch, React Query & SWR queryFns | express — app.use() mount graph, cross-file |
angular-http — Angular HttpClient | fastify — verb shorthand + route({}) |
next — file-based (app router + pages API) | |
fastapi — APIRouter prefix + include_router | |
flask — blueprints + methods=[] | |
spring — @RequestMapping + @*Mapping | |
openapi — universal: any OpenAPI 2/3 doc (file or URL) |
That covers MERN (rtk/axios ↔ express), MEAN (angular ↔ express), MEVN (axios ↔ express), Next full-stack, Python (fastapi/flask), and enterprise (angular ↔ spring) — plus openapi for any backend that emits a spec (Express+swagger-jsdoc, FastAPI, Spring springdoc, .NET Swashbuckle, …).
Notes:
app.use() mount graph across require/import boundaries and nested routers; routers it can't reach are flagged, never dropped.GET/POST/…; pages-router handlers serve any verb (matched as ALL). `users/${id}?x=${q}` ) are surfaced as unverifiable rather than guessed.tieline.config.json — repo paths resolve relative to the config file.
{
"client": { "adapter": "rtk-query", // rtk-query | axios-fetch | angular-http "repo": "../web", "roots": ["src/redux/apis"], // dirs to scan for call sites "basePath": "/api/v1" // stripped from call sites before matching }, "server": { "adapter": "nestjs", // nestjs | express | fastify | next | fastapi | flask | spring | openapi "repo": "../api", "roots": ["src"], "globalPrefix": "api/v1", // stripped from routes before matching "spec": "openapi.json" // openapi adapter & doctor only — file path or URL }, "ignore": ["internal/.*"], // regexes on the normalized path "failOn": ["drift"] // buckets that make check exit non-zero }
tieline check --html report.html writes one self-contained file (inline CSS/JS, no external assets) you can open in any browser or attach to a PR:
∅ no route node catching calls that land nowhere; hover a resource to highlight its linkstieline doctor — does your code match your published docs?doctor diffs routes parsed from source (a native adapter like nestjs) against the routes declared in your OpenAPI spec (server.spec):
tieline · doctor code (nestjs) ↔ spec (http://localhost:9999/doc-json)
❌ 4 undocumented (in code, missing from the published spec) GET /billing/invoices src/billing/billing.controller.ts:54 POST /webhooks/stripe src/webhooks/webhooks.controller.ts:21 ... 👻 1 phantom (in the spec, no matching route in code)
✅ 312 agree ❌ 4 undocumented 👻 1 phantom (316 code routes, 313 spec routes)
Run it alongside check and your spec can never silently drift from your code.
tieline ships an MCP server so an agent can ask "do these two repos still agree?" and get the same structured result the CLI produces — no LLM does the analysis, the deterministic engine does. It's still zero-dependency: the server is hand-rolled stdio JSON-RPC, no SDK.
Register it with any MCP client (Claude Code, Claude Desktop, …):
{
"mcpServers": { "tieline": { "command": "npx", "args": ["-y", "-p", "@nugehs/tieline", "tieline-mcp"] } } }
Or, from a global install (npm i -g @nugehs/tieline), just "command": "tieline-mcp".
It exposes five tools, each returning JSON:
| Tool | Args | Returns |
|---|---|---|
tieline_check | config? | totals + drift + unverifiable (the drift gate) |
tieline_list | config? | the full resolved contract map |
tieline_orphans | config? | backend routes no frontend call reaches |
tieline_doctor | config? | undocumented + phantom (code ↔ spec) |
tieline_init | cwd? | auto-detect nearby repos, write a config |
config defaults to searching upward from the server's working directory, exactly like the CLI — so an agent dropped into a repo with a tieline.config.json can just
Nur der Anfang wird gezeigt; den Rest findest du im Repo