TestWatch vs Playwright Test Agents
Playwright’s Test Agents are three AI agents that plan, write and repair tests inside your repo, driven by your coding assistant. TestWatch writes the same kind of Playwright test as a hosted service, then runs it, explains failures and proposes fixes you approve.
Checked 17 October 2026. Facts about Test Agents come from playwright.dev and are linked. Spotted something out of date? Tell us and we fix it.
The short version
- Use Playwright Test Agents if
You want it free and local, the tests should live in your repo from the first commit, and someone on your team is comfortable running the suite in CI, reading failures and reviewing what the healer changed. You also keep every browser Playwright supports, Firefox included.
- Use TestWatch if
Nobody owns end-to-end tests and you want them to run anyway: on a schedule or from CI, with a reason for each failure and a fix you approve when the page changed. You can still take the code with you: the whole suite exports as a Playwright project.
Side by side
| Playwright Test Agents | TestWatch | |
|---|---|---|
| What it is | Three agents: a planner writes a Markdown test plan, a generator turns it into test files, a healer repairs failing tests | A hosted service: you describe a flow, an AI agent writes and verifies the Playwright test, and TestWatch runs it |
| Where it runs | In your repo, from a coding assistant: VS Code, Claude Code, Codex or opencode (npx playwright init-agents) | In our browsers: desktop and mobile Chrome, mobile and desktop Safari (emulated) |
| Who maintains it | You. Agent definitions should be regenerated whenever Playwright is updated | We run the service; you review each new test and each proposed fix |
| Cost | Playwright is free and open source (Apache-2.0); you pay for your coding assistant’s AI usage and your CI minutes | Free plan, then from $99 a month, priced by tests with runs included. Passing runs make no AI calls |
| Scheduling and CI | Whatever you set up: they are ordinary Playwright tests in your repo | Schedules on every plan; GitHub Action, CLI, REST API and deploy hooks on Starter and up |
| When a test fails | The healer replays the failing steps, inspects the UI and patches the test, or marks it skipped if it believes the functionality is broken | A verdict with evidence: page changed, real bug, flow changed, flaky, CAPTCHA, login unavailable or app down |
| Fix approval | The healer edits the test files; you review the change like any other change in your repo | A proposed fix to locators and waits, applied only when you approve it. Fixes don’t change what a test checks, except a check’s expected text on a pull-request preview, which you confirm when you accept the fix. An accepted PR fix is applied to the main test once it passes there twice, with a notice and Undo |
| Code ownership | Plans in specs/ and tests in tests/, in your repo | Readable Playwright code per test; export the whole suite as a runnable Playwright project, on every plan |
| Logins, 2FA, email | Not covered on the Test Agents page; a seed test you write sets up the starting state | Test accounts with encrypted credentials, authenticator-app (TOTP) codes, and built-in test inboxes for emailed links and codes. No SMS codes or Google/GitHub/SSO sign-in |
The real difference: who owns the loop
Writing the test is the easy half. Test Agents do it well, in your repo, and the generator checks its selectors and assertions against the live app as it goes. The harder half is everything after: running the suite every day, noticing when it fails, telling a renamed button from a broken checkout, and deciding whether to change the test or fix the app.
With Test Agents that loop belongs to you and your CI. The healer helps with the repair step, and its docs say it may mark a test skipped when it believes the functionality is broken, so someone should read what it did. With TestWatch the loop is the product: runs on a schedule or from CI, a verdict for each failure, and a fix that waits for your click.
Using both
They aren’t exclusive. Developers can keep writing tests in the repo with Test Agents while TestWatch watches signup, login and checkout on production. And because a TestWatch suite exports as a standard Playwright project, you can move those tests into the same repo later.
How we compared
We read Playwright’s Test Agents, codegen and MCP docs on 17 October 2026 and linked each fact to its page. Where the docs don’t say something, the table says so instead of guessing.