Loop QA automatically QAs web apps. Give it a URL and it will start exploring, then map out and test the different user journeys. It records while testing and analyzes those recordings to produce a feed of detailed, actionable bug reports. You can paste these into your agent to fix the issue, or pipe them to your software factory to take care of them as soon as they’re reported.
Here is an example project for a hiring platform that Claude Code (Opus 4.6) built and claimed was completely working. Loop QA identified numerous problems, including the following:
- Application submission hangs at the final step.
- Searching candidates by full name doesn’t work.
- The landing page has redundant CTAs “View Open Positions” and “Browse Jobs”.
- The main dashboard cards are not aligned and show redundant counts.
Each bug report has a detailed analysis identifying the root cause. Loop QA uses Replay MCP to gather evidence from its recordings and calls a judge model to make sure the evidence supports the conclusions and that the bug is being described completely.
Loop QA’s goal is to accurately explain as much as it can of the real problems standing in the way of user acceptance: not just broken functionality but also poor performance, confusing interfaces, and glitchy behavior. It is open source, agent-built, and free to use at low volume.
Try it out and let us know what you think! If you’re building apps with AI at scale and want to make them better, get in touch.
