PROJECT CASE STUDY
Website-building agent team
Six agents coordinate planning, content, implementation and checks.
Prototype
Page updated:
CrewAI · Python · Flask
What I built
The application turns a user’s business, style and page preferences into a structured brief. Six agents handle requirements, engineering planning, copy, frontend implementation, testing and review.
Why generate files individually?
A single long response can leave the final files incomplete. Producing files sequentially helps keep them complete and consistent, at the cost of more context and model calls.
Checks outside the model
Deterministic Python checks inspect required HTML information, internal links, missing files and placeholder content. A positive model review cannot override a critical deterministic failure.
Delivery and limits
Flask presents the preference flow and generation progress. The project is a prototype. Generated websites still require browser checks and human review; the interface screenshots are not proof that every generation succeeds.
Next objective
Expand browser checks for generated sites and document failure cases.