Definition: A QA agent is an AI agent that tests software separately from the agent that wrote it. It runs logic and unit tests, plus UI and flow tests that click through screens, buttons and whole user journeys, and it decides whether a task can be marked done. Its purpose is simple: an AI shouldn't grade its own homework.
What a QA agent checks
- Logic: do functions return the right results for normal and edge-case inputs?
- UI: do screens render, and do buttons and forms work?
- Flows: can a user complete a whole journey, such as sign up, book, pay and get a confirmation?
- Regressions: does everything that worked before still work?
Example from a real workflow
In Devfab, the QA agent is a switch in each workspace's controls: QA agent · review before Done. When it's on, every ticket passes QA review before it can be marked done. That's safer but slower. When it's off, tickets can skip QA, which some teams use for quick experiments.
Related terms
FAQ
Can a QA agent replace human testers?
It handles the repetitive checks on every change. People still decide what matters, review test reports and judge usability.
Why keep QA separate from the coding agent?
The agent that wrote the code makes the same assumptions when it tests it. A separate check catches what those assumptions hide.
Read next
Frequently asked questions
Can a QA agent replace human testers?
It handles the repetitive checks on every change. People still decide what matters, review test reports and judge usability.
Why keep QA separate from the coding agent?
The agent that wrote the code makes the same assumptions when it tests it. A separate check catches what those assumptions hide.
