TestDriver is an AI-powered end-to-end UI testing tool, positioned as an AI quality engineer for every pull request. It reviews every pull request by running the application in a real desktop sandbox and testing it like a real user, rather than just reviewing code or driving a browser.
Key capabilities:
- Automatically generate tests and catch regressions: Mention @testdriverai in any GitHub repository, and TestDriver will write UI tests and catch regressions before merge.
- AI visual verification of pull requests: The visual agent reads the diff to understand the changes, then generates end-to-end tests directly from the PR content, proving that the new behavior works by clicking, typing, and reading the screen. No selectors, DOM, or setup required.
- Contribute tests: Once the agent has a passing test, it opens a new pull request and commits the generated test code to the repository, with no handwritten boilerplate required. Tests are pure Vitest and run consistently locally and in CI, executing across web, desktop, and extensions on real virtual machines.
- Self-healing capability: When UI changes cause cached elements to no longer match, TestDriver re-invokes AI to find the elements, updates the cache, and continues running. Tests can survive redesigns without breaking on every pixel change and without needing to be rewritten.
- Debug failures: Jump directly from a red build to video playback of the failed run and view network logs and the sequence of every click.
- Inspect network calls: Inspect HTTP requests and responses in detail, view headers, payloads, and timing, and quickly diagnose API issues and failed interactions.
- Performance monitoring: Monitor resource usage during test execution and catch memory leaks and performance bottlenecks.
- Step-by-step execution history: Track the complete timeline of every click, input, and assertion to pinpoint exactly where the test diverges from expected behavior.
- Prevent regressions: Track pass rates, flaky tests, and failure trends for each run so regressions can be discovered as soon as they appear. Export to JUnit XML for any existing reporting tool.
How to use it: Mention @testdriverai in a GitHub repository to trigger it. TestDriver automatically runs every pull request in a real desktop sandbox, generates tests, and submits them as a new pull request. The generated tests are in pure Vitest format and can be integrated into existing pipelines.
Use cases: Suitable for engineering teams that want to automatically perform end-to-end UI testing during the pull request stage across web, desktop, and extension platforms. Especially suitable for teams that want toreduce manual test writing, avoid fragile tests caused by UI changes, and quickly debug failed tests.
Limitations and billing: Starts at $20 per seat per month, including 120 minutes of testing time. Seats are automatically billed for each GitHub user who receives a PR review or runs tests. Test minutes are billed by sandbox runtime at $0.14 per minute, occurring automatically when PRs are generated and tests are run. An unlimited testing option is available with no per-minute fee and requires a shared team license. Enhanced security and customization, expert setup, and support are also available. Specific limitations and further details are not provided at this time.
