Builder can provide feedback directly on pull requests whenever a Builder user submits a request. Team and Enterprise accounts can enable this feature for specific Projects.
Builder's code review and quality assurance agents only run for pull requests created by Builder users.
To enable the code review agent for a Project:
- Go to your Projects page.
- Click the three-dot menu next to a Project.
- Go to the Review tab.
- Find the Quality Review heading, and click the toggle next to Code Review.
After toggling on the Code Review option, provide details to the review agent in the Custom Instructions section on how it should review code, or what it should focus on.
This information can also be added to an AGENTS.md file.
Code review agents consume credits, just like other AI features.
By default, Builder's review agent categorizes issues by critical and medium priority. The review agent also highlights the pull request's positive qualities.
In the video below, the review agent has found two critical issues and two issues of medium priority.
The quality assurance agent tests the application in a browser to verify implementations are correct.
This agent runs alongside the code review agent, programmatically checking interactive elements of the application specified by either the pull request or by your own custom instructions.
To enable the quality ensurance agent for a Project:
- Go to your Projects page.
- Click the three-dot menu next to a Project.
- Go to the Review tab.
- Find the Quality Review heading, and click the toggle next to Quality Assurance. The Code Review option must also be enabled.
After toggling on the Quality Assurance option, provide optional login details to the agent. If your application requires a login to reach important features, this step is critical.
You can also add details to the review agent in the Custom Instructions section on how it can interact with your application, or what features it should focus on.
Quality assurance is not available to users solely using the Builder CLI.
Builder's AI runs interactive tests based on your pull request and the custom instructions provided within the Project settings.
In the video below, the user scrolls through a pull request that contains several details and reviews from Builder. The last entry displays the quality assurance tests.
Quality assurance agents consume credits, just like other AI features.
Quality assurance results come with a Fix in Builder button. When clicked, this button brings you to a Builder pull request page where you can select which specific issues to fix.
In the video below, the Fix in Builder button is clicked. The quality assurance comment is selected, and then the Verify and fix all button is clicked. Changes are made to the original Builder Project branch, which can then be pushed to the same pull request.
Builder's quality assurance agent captures video of its live tests, so you can see what it saw during testing. Clicking the video on the pull request brings you back to Builder and its video viewer.
Within the video viewer, you can:
- Watch tests run by the agent live in a video format
- Review specific tests run by the agent
- Read comments from the agent on each test case
- Review network and console errors that occurred during the test run