Generate initial code

Every file in the scaffold starts as a skeleton: the right place in the project, with the body left to write. Generate code writes those bodies — working starter code for your application, produced one file at a time.

Where it lives

In the Export tab's Scaffolding panel, the Build order row sits beneath the file tree. Expand it to see each file as a numbered step, with a running count of how many are done. Generate code starts a run; Stop ends one between files. Selecting a step opens that file in the preview.

Files are generated in the order the project needs them — shared pieces before the pages that use them — so every file fits the ones written before it.

What informs each file

Generation follows the plan. The project description, roles, and features from Domain, the entities from Data, and your implementation instructions shape every file. A page also respects the role access recorded in Security, and is written toward the features Traceability links to it. The more of the plan you have filled in, the more the generated code is about your application rather than a generic one.

Cost and batches

Each file costs 1 AI credit, charged only for the files a run actually writes. A run covers up to 20 files and says how many remain; pressing Generate code again continues from the first unwritten file, so a large project is generated in passes. Finished files are never re-generated — and never re-charged. A file that fails is marked in the list and can be retried on the next press.

Reviewing and keeping the code

A generated file replaces its skeleton in the preview, and anything left for you to decide — a guess the generator made, an integration it stubbed — appears as a note above the code. Discard code removes every generated file after a confirmation, since re-generating them costs credits again.

Generated code is kept for the current session only. Download the .zip or push to GitHub to keep it — both contain exactly the files the preview shows.