What the agent does
The agent reads the current test setup before replacing it. In a project like chibat/chrome-extension-typescript-starter it finds jest.config.js with roots: ["src"] and a transform that runs ts-jest against tsconfig.test.json, which extends tsconfig.json only to switch moduleResolution from bundler to node. package.json has "test": "npx jest", with jest 29, ts-jest 29 and @types/jest as dev dependencies next to webpack 5, ts-loader, copy-webpack-plugin, prettier and @types/chrome. The only test is src/__tests__/sum.ts, two cases against src/sum.ts using the global test and expect.
It reads the rest of the extension to know what the tests will need to touch next. public/manifest.json is Manifest V3 with a background.service_worker, a popup, an options page and a content script on <all_urls>; src/popup.tsx calls chrome.action.setBadgeText, chrome.tabs.query and chrome.tabs.sendMessage; webpack/webpack.common.js bundles popup, options, background and content_script entries with ts-loader. .github/workflows/build.yml runs npm ci, npm run build --if-present and npm test on Node 20 and 22, so the migrated suite runs in CI without a workflow change.
The agent installs vitest and removes jest, ts-jest and @types/jest, adds a vitest.config.ts with test.include set to the existing src/__tests__ pattern and globals: true so test and expect keep working unimported, and changes the test script to vitest run. It deletes jest.config.js and tsconfig.test.json, because Vitest uses esbuild and reads tsconfig.json as it is, and adds vitest/globals to the types in tsconfig.json in place of the Jest types. Since src/popup.tsx uses the chrome namespace, it adds a src/__tests__/setup.ts registered as setupFiles that stubs chrome.action, chrome.tabs, chrome.runtime and chrome.storage with vi.fn(), so the next test written against the popup or the background worker has a mock ready, and writes one test using it so the setup is exercised.
Before pushing it runs npm run build to confirm webpack and ts-loader still compile every entry with the changed tsconfig.json, then npm test and npx vitest run --coverage to see the two original cases and the new one pass. It opens the pull request, reads the build workflow on both Node versions, and fixes anything that fails there.
Start it with one API call
POST https://api.phoenix.so/o/{org}/sessions with this body:
{
"agent": "claude",
"input": "Migrate this Chrome extension's tests from Jest to Vitest. Read jest.config.js, tsconfig.json, tsconfig.test.json, package.json, the tests under src and the GitHub Actions workflow first. Install vitest, remove jest, ts-jest and @types/jest, add vitest.config.ts with globals enabled and the existing test pattern, change the test script to vitest run, delete the Jest config and the test tsconfig, and replace the Jest types with vitest/globals in tsconfig.json. Add a setup file that stubs the chrome namespace APIs the extension uses and one test that exercises it. Run the webpack build and the tests before you push. Do not change the manifest or the webpack config. Open a pull request and make sure the build workflow passes on it.",
"resources": [
{
"url": "https://github.com/your-org/your-app"
}
]
}- Get an API key. In the console, open Settings, then API keys, and create a key for your org. Send it in the Authorization header as a Bearer token.
- Send the request. POST the body in request.json to https://api.phoenix.so/o/{org}/sessions, with your repository as the resources URL. The run starts in its own cloud sandbox.
- Read what came back. Follow the run on its live stream or wait for the run.completed webhook, then read the result and its trajectory.
What comes back
What comes back is a pull request on a new branch that removes jest.config.js and tsconfig.test.json, adds vitest.config.ts and src/__tests__/setup.ts, changes package.json and package-lock.json to swap the Jest packages for vitest, updates tsconfig.json types, keeps src/__tests__/sum.ts passing without edits to its assertions, and adds one test that uses the chrome stubs. In a project shaped like the TypeScript starter, public/manifest.json and the webpack/ directory are untouched, and the build workflow runs unchanged because its npm test step now runs Vitest.
The proof is that workflow on the pull request, green on both Node versions in its matrix, plus the npm run build, vitest run and coverage output kept in the session's trajectory with every command. The pull request description lists the packages removed and added, the config files deleted, and which chrome APIs the setup file stubs.
Before merging, a reviewer checks that the stubbed chrome APIs are the ones their real tests will call, that globals: true is the style they want rather than explicit imports from vitest, and that nothing else in the repository still imports from @jest/globals. Then they can write popup and background tests against the mock without further setup.
Run it on every extension repository still on Jest
Run this once per repository: have a script send the same request with each repository's URL in resources and a fresh UUID as request_id, so a retried request returns the run already started instead of a second pull request. Each run has its own sandbox and its own branch, and the Rush GitHub App must be installed on each repository; GET /o/{org}/github lists the ones it can clone.
Set a webhook URL on the API key and Rush POSTs run.completed or run.failed when each run's turn finishes, with session_url so your server can post the pull request link to the team's channel. A batch across many extensions needs no queue on your server: with max_concurrent runs in progress, 4 by default and up to 100, the key answers further requests 202 with status: "queued" and starts them, oldest first, when a slot frees.
The use case this job follows, with its automation and trigger: Hand off a long task.
Why run it on Rush?
A test-runner migration is mostly bookkeeping until a type setting breaks the webpack build or a chrome call explodes in the first real test. On Rush the agent does the swap in its own sandbox, runs the webpack build and the suite there before the first push, and then reads the build workflow on its own pull request across the Node matrix. Your laptop never installs a second test runner alongside the first.
The result is checkable rather than claimed. The workflow on the agent's pull request shows the suite passing under Vitest on both Node versions, and the trajectory keeps the build and test output, so a reviewer can see the two original cases and the new one pass before opening the diff. Because each run is independent, the same request can migrate every extension repository the team maintains at once, one pull request each.
Frequently asked questions
Will it change my webpack config or manifest?
No. The prompt tells it not to, and Vitest runs the TypeScript sources directly through esbuild without ts-loader, so webpack/webpack.common.js, webpack.dev.js, webpack.prod.js and public/manifest.json are left alone. The only build-related edit is the types entry in tsconfig.json, and the agent runs npm run build to show webpack still compiles.
Do my existing tests need rewriting?
Tests that use the global test, expect and describe run unchanged with globals: true, which is how src/__tests__/sum.ts is written. Tests that import jest.fn or jest.mock are changed to vi.fn and vi.mock, and the agent lists each such edit in the pull request. Snapshot files keep working.
How are chrome.* APIs handled in tests?
A setupFiles entry registers src/__tests__/setup.ts, which defines a chrome global with vi.fn() stubs for action, tabs, runtime and storage, the namespaces src/popup.tsx and the background worker use. One test exercises the setup so it is not dead code. Ask for sinon-chrome or a fuller mock package in the prompt and the agent uses that instead.
Why delete tsconfig.test.json?
It existed only to switch moduleResolution from bundler to node for ts-jest. Vitest's esbuild transform does not need it and reads tsconfig.json directly, so keeping the file would leave a config nothing uses. The pull request description says so, and a reviewer can keep it if another tool depends on it.