Agentic testing with Playwright: how AI agents plan, write and repair tests
How AI agents do QA work with Playwright today: the loop, the tools (test agents, MCP, Agent SDK), where agents go wrong and the guardrails that work.
Agentic QA
Agentic QA, with a person who signs the release
In agentic testing, AI agents take on whole QA tasks: planning tests, writing them, running them and repairing them. We run it as a service for your web, API and mobile app, and a human QA engineer reviews everything the agents produce.
What agentic QA means
Agentic QA gives the QA work itself to AI agents that can read, plan, write code and run it. An agent reads your specs, tickets and code, drafts a test plan, writes the tests, runs the suites and collects evidence for each failure. When the product changes, it proposes fixes for the tests that broke.
Our agents are built with the Claude Agent SDK and run on Claude. The tests they write are plain Playwright, Appium, Maestro or Mobilewright code in your repository, so your team can read them and keep them.
Agentic QA, agentic testing, AI QA: the names vary. The question that matters is who checks the agents' work. Here, a QA engineer does, on every change.
Who does what
The agents do the volume, a person makes the call
Claude agents
- Draft test plans from your specs, tickets and code
- Write new tests and update tests the product has outgrown
- Run the suites and collect evidence for each failure
- Propose fixes for broken tests
The QA engineer
- Sets the strategy and the priorities with you
- Reviews every change before it is merged
- Confirms each bug and its severity
- Signs the Go or No-Go call before each release
Why a person stays in the loop
An agent asked to fix a failing test can make it pass by checking less. So an automatic check rejects any fix that would weaken what a test asserts, and a person reviews each fix before it is merged. A green run should mean the product works, not that the tests stopped looking.
A person also looks at every failure before it becomes a bug report or a test fix, and signs the Go or No-Go call before each release. Shipping stays your decision.
AI testing for web, API and mobile
Web with Playwright, API tests, and mobile with Appium, Maestro or Mobilewright: we choose the best fit for your app. The tests run in your own CI on every pull request, with no fees per run.
What it costs
US$2,000 a month for up to 500 user flows. Test runs, AI credits and hours are never billed. New clients start with a 60-day pilot at half price. The pricing page has the calculator and every term.
See agentic QA on real code
agentic-qa is our public sample project: Claude agents doing the QA work on a demo shop we built, with a live Go/No-Go report published from CI. It shows how the work is done, not results for a client.
Your AI testing tool closed down? Octomind, Bugster and DogQ alternatives covers what to save and where to move your tests.
Guides
Guides on agentic testing
Self-healing tests: what AI should fix and what it must never touch
What self-healing tests really change, which AI fixes are safe, which need a person, and a review workflow, with a before/after Playwright example.
Flaky tests in Playwright: causes, fixes and a quarantine policy
A field guide to flaky tests in Playwright: what causes them, how to find them with repeats, retries and traces, how to fix each cause, and how to quarantine safely.
Tell us what you are shipping next
One email is enough to start: your product, your platforms and how often you release. You get a plain answer on fit and, if it fits, a written proposal for the 60-day pilot with a short video of how we would test your product.
Or write to qa@releaseworthy.app. Still deciding after the proposal? We book a Google Meet call for your questions.