Find gaps in your API mock coverage before they reach CI.

MSW handlers drift. API calls get added without mocks, and old mocks stay behind after the code moves on. msw-inspector scans both sides, compares them, and reports what is covered, what is not, and what looks stale.

npm install -D msw-inspector-cli

The npm package is published as msw-inspector-cli because the original msw-inspector name is already taken on the registry. The installed binary remains msw-inspector.

Run it from the project root:

npx msw-inspector

Or run it without installing first:

npx msw-inspector-cli

Useful flags:

npx msw-inspector \
  --handlers "src/**/*.{ts,tsx,js,jsx,mts,mjs,cjs}" \
  --sources "src/**/*.{ts,tsx,js,jsx,mts,mjs,cjs}" \
  --exclude "**/dist/**" "**/*.d.ts" \
  --base-url "https://api.example.com" \
  --report-file msw-inspector.json \
  --format text

The CLI prints a human-readable summary by default. Use --format json when you want the full report for CI or a downstream action.

If your app uses relative URLs but you want origin-aware matching, set --base-url. That resolves relative handlers and API calls against one canonical origin, which is useful when the same pathname exists on multiple backends.

Text output looks like this:

✓ 23 handlers found
✓ 31 API calls found
✗ 8 unmocked endpoints
✗ 3 stale mocks

Coverage: 74% (23/31)

Here is a real run against typejung.com:

The JSON report written by --report-file includes:

{
  "schemaVersion": 1,
  "summary": {
    "mockedCalls": 23,
    "totalCalls": 31,
    "usedHandlers": 20,
    "totalHandlers": 23,
    "staleHandlers": 3,
    "unmockedCalls": 8,
    "percentage": 74.2
  }
}

Dogfood run on typejung.com:

{
  "summary": {
    "mockedCalls": 0,
    "totalCalls": 24,
    "usedHandlers": 0,
    "totalHandlers": 0,
    "staleHandlers": 0,
    "unmockedCalls": 24,
    "percentage": 0
  },
  "unsupported": 7,
  "sampleUnmocked": [
    "POST https://oauth2.googleapis.com/token",
    "GET https://www.googleapis.com/oauth2/v2/userinfo",
    "POST /api/chat",
    "POST /api/create-checkout-session"
  ]
}

That run surfaced a complete mock gap across auth, billing, and AI endpoints instead of a single missing handler.

I ran the analyzer against three real repositories:

• typejung.com: 0 handlers, 24 API calls, 24 unmocked endpoints, 7 unsupported dynamic patterns.
• msw: narrowed to the browser test slice, 191 handlers, 2 API calls, 190 stale mocks, 28 unsupported patterns.
• oss-msw: same slice, 191 handlers, 2 API calls, 189 stale mocks, 28 unsupported patterns.

The strongest product signal came from typejung.com: it immediately showed that a real app could have a non-trivial API surface with zero MSW coverage. The two MSW repos exercised the other side of the problem, where handlers accumulate and drift stale when the active request surface gets narrower.

The first release is intentionally narrow:

• msw http.* handlers
• legacy msw rest.* handlers
• handler matchers from string literals, static template literals, static consts, new URL(...).href, new URL(...).toString(), and String(new URL(...))
• fetch(...), window.fetch(...), globalThis.fetch(...)
• common axios call shapes, including axios.get(...), axios.request(...), axios(...), and same-file axios.create(...) instances

The CLI ships with a matching GitHub Action wrapper in a separate repository: felmonon/msw-inspector-action. It reads the JSON report that the CLI already produced, writes a job summary, and can optionally upsert one sticky PR comment.

Marketplace listing: MSW Inspector

name: msw coverage

on:
  pull_request:
  push:

jobs:
  inspect:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with:
          node-version: 20
          cache: npm
      - run: npm ci
      - run: npx msw-inspector --report-file msw-inspector.json --format json
      - uses: felmonon/msw-inspector-action@v1
        with:
          summary-file: msw-inspector.json
          comment: true

The action does not compute a baseline delta yet. It publishes the current report cleanly and predictably.

• It does not try to infer custom wrapper helpers.
• It does not resolve cross-file constants or imported axios instances.
• It does not analyze GraphQL, WebSocket, or SSE handlers.
• It reports dynamic or ambiguous patterns as unsupported instead of guessing.

npm install
npm test
npm run typecheck
npm run build

If you are changing the scanning logic, keep the test fixtures small and explicit. The tool is more useful when it stays opinionated.