How a build works

LoopCodeLab turns your idea into a plan of stories. Agents can build those stories in parallel. A master agent reviews each piece, merges accepted work, and runs a final check before delivery.

The path to delivery

1. You describe the result you want.
2. The planner turns it into stories.
3. Workers build the stories.
4. The master agent reviews and integrates the accepted work.
5. A final pass checks that the pieces fit together.
6. LoopCodeLab delivers the live preview and repository, plus any app or media artifacts for the selected output.

Stories are editable units of work

A story is one defined piece of the build. It has a title, a brief, and acceptance criteria. You can edit or remove stories during review, before the build starts.

When something goes wrong

The build system watches for stalled work and can nudge or retry it. It can identify a bad provider key as an account problem and route the story to another agent when possible. If a build fails, use Doctor to diagnose the problem again and resume.