Introduction
TryCase is an AI QA agent: every time a pull request is submitted, it actually runs the application, measures the user path affected by the change, and then sends the conclusion, subtitled video and screenshots back to GitHub.
No need to write tests first. It reads the PR changes itself, determines which user flows will be affected, and then clicks through them just like a real person would. In addition to the web interface, API, CLI, workers, and packages can also be run through real commands. Monorepo support - scan the applications in the warehouse when setting up for the first time, and reuse them every time.
The triggering method is optional: run every push, or comment /trycase run in the PR. The GitHub App is installed, and both public and private repositories are acceptable.
A cost-friendly design is that you can bring your own AI subscription: by connecting to the existing ChatGPT subscription or OpenRouter key, the AI quota will no longer be included in the package. The package is only calculated based on the test duration. Officially, Claude subscription support is being worked on. It is suitable for small teams that make frequent changes and never have time to complete end-to-end testing.
