Replay QA Promises Root-Cause Bug Reports for Any Web App, Tested on AI-Built Apps
Replay has introduced Replay QA, a drop-in tool that the company says can find and explain every bug in a web application. Replay set out the product in a blog post dated 3 November 2026. It presents the tool as an extension of its time-travel debugger for Chromium and Node.js, the browser and server-side JavaScript runtimes respectively.
Replay states its mission as making it possible to reliably and automatically detect and fix every bug users encounter on the web. The debugger underneath is the company's core technology. According to Replay, its forked runtimes capture non-deterministic inputs, meaning inputs that can differ from run to run, with low overhead. That allows a recorded session to be replayed deterministically, producing the same result every time. Replay QA draws on that recording layer to do more than flag a failure.
Replay says Replay QA can explore and test an application on its own or with guidance. Starting from a URL, it can map out the application's user journeys, write tests for them and capture runtime recordings of each session. The company says the tool autonomously tests every build for bugs, security flaws and accessibility failures, and files root-caused bug reports that teams can act on.
The defect categories Replay lists are broken features, glitches, performance issues and security holes. For each bug, Replay says Replay QA identifies the root cause and supplies evidence from the recording. The company also describes self-healing capabilities, under which Replay QA creates recordings by reproducing user sessions from partial and redacted data, meaning data with sensitive details removed.
To test the product, Replay says it ran Replay QA against 40 apps promoted by AI app-building platforms including Lovable, Replit, Base44 and Bolt.new. After confirmation and deduplication (removing repeat reports of the same bug) by Astra, the company reports finding over 900 bugs across those applications. Replay positions the tool for builders of AI-generated apps, saying it runs real browser tests on such apps, records every session and gives builders proof that their app works before launch.
The broader context here is a shift in where quality assurance sits in the development loop. Platforms that generate a working front end from a prompt compress the time between idea and deployment, and the testing burden moves with it. A conventional test suite written by hand covers the paths its authors anticipated. A tool that maps journeys from a URL and records every session is attempting to cover the paths they did not.
In my view, the phrase "every bug" is a statement of mission rather than a measured result. That sample was drawn from applications promoted by the builder platforms themselves, so the more informative test will be how the tool performs on applications nobody selected for showcasing. Confirmation and deduplication also shape any bug count.
The practical value is easier to state. If the root-cause evidence holds up under independent use, a developer receives a reproduction and a cause rather than a ticket reading that the checkout button is broken. That difference, multiplied across the volume of apps these platforms now produce, is the outcome worth watching in the months ahead.
