Updated
Vibe coding

AI Testing

Let AI agents write, run and maintain end-to-end tests of your app, so agent-written code gets clicked through before it ships, without you writing the tests by hand.

Ask your AI about this, with this page as the source:ChatGPT ↗Claude ↗Perplexity ↗

Trends

Public counts each month: pull requests on GitHub, package downloads, stars, Product Hunt followers. They show direction, not market share.

ToolGitHub PRs, Sep 2026DownloadsStarsProduct Hunt
TestDriver146 reviewed18 opened20.7k/wk−19% vs npm★ 243—
TestSprite23 reviewed11.9k/wk—4.9k followers
Playwright—92.2M/wk+17% vs npm★ 97.5k—
Meticulous—96.5k/wk−20% vs npm—8 followers
OctomindArchived————

The real choice

Tests you own, or tests a service runs for you. Some tools write real test files — Playwright, Vitest, YAML steps — into your repo, where they run in your CI and stay yours if you leave; others test your deployed app from the outside on every pull request and report what broke, with nothing to maintain but nothing to take with you. Beneath both, a browser MCP server lets your own coding agent click through the app as it works, which is checking, not a test suite.

Pick by situation

Tap the ones that are you — the tools that fit light up below.
  • You want your coding agent to click through the app it just changed, at no extra costPlaywright
  • You want AI-written end-to-end tests stored in your repo and run in your CIMomenticTestDriver
  • You want tests run on every pull request without writing or maintaining any yourselfTestSpriteQA Wolf
  • Your frontend is complex and regressions slip through reviewMeticulous

Compare the tools

Grouped by the side of the choice they answer, not ranked. Open a row for when to use it, the trade-off and what makers say.

ToolBest for
Browser for your agent1
Playwrightframework · open sourceMicrosoft's open-source MCP server that lets your coding agent drive a real browser through Playwright's accessibility tree, to check its own changes or turn a session into Playwright tests.No evidence yet92.2M/wk+17% vs npm

Use it whenYou already use a coding agent and want it to verify the app in a browser before you review.

Trade-offThe agent checks; it doesn't leave a suite behind unless you ask it to write tests, and accessibility snapshots use a lot of context.

No maker's product we track shows it for this yet.

Tests in your repo1
Momenticservice · free tierEnd-to-end tests for web, iOS and Android written as YAML steps and AI actions in your repo, run from a CLI locally, in CI or on hosted browsers and emulators, with an MCP server for coding agents.No evidence yet

Use it whenYou want AI to write and heal tests that still live in your repo and run in your CI.

Trade-offBilled in credits; the free plan stops at 2,000 credits a month and paid use starts at $125 a month.

No maker's product we track shows it for this yet.

Hosted test agent2
QA WolfserviceAgents that map your app and write, run and fix end-to-end tests in Playwright and Appium for web, mobile and desktop apps, billed per test run, with human QA services available.No evidence yet

Use it whenYou want tests maintained for you with no seat fees, and may want people to own QA later.

Trade-offPer-run pricing ($1 per web run) grows with the size of the suite and how often it runs.

No maker's product we track shows it for this yet.

TestSpriteservice · free tierAn agent that plans, writes and runs frontend and backend end-to-end tests against your app and reports what broke, from a web app, a CLI, an MCP server for coding agents, or automatically on GitHub pull requests.No evidence yet11.9k/wk

Use it whenYou want pull request testing with a free plan to start, and your coding agent able to trigger tests itself.

Trade-offCredit-based plans, and the free plan covers pull request testing on one repository.

No maker's product we track shows it for this yet.

Before you choose

How to approach it

Start with your coding agent checking its own work in a real browser through Playwright MCP, and add a test suite once the same flows keep breaking. Cover the three or four flows that make money — sign-up, the core action, checkout — before anything else. Run tests on every pull request against a preview deployment, and read the failures yourself: an agent asked to make a test pass may change the test instead of the code.

Common mistakes
  • Letting an agent "fix" a failing test by weakening the assertion, so the suite goes green while the bug ships.
  • Aiming AI tests at every page at once, then drowning in flaky failures nobody has time to triage.
  • Running tests against production with real payment or email providers, so every run sends real emails or charges.

Other options

Real choices most makers here won't need to weigh.