CLAUDE.md ships at the project root — Claude Code reads it on every session automatically.
3-layer pattern, Riverpod conventions, folder structure — every rule encoded for the AI.
Plain Markdown — fork, scope, layer your team conventions on top of the defaults.
Without explicit rules, AI assistants default to generic Flutter patterns. With them, every suggestion follows your conventions — saving review time, preventing drift, and onboarding both humans and AIs from the same source of truth.
Stop fixing generic code
AI starts from your conventions, not Stack Overflow defaults. Less re-prompting, less rewriting.
Consistency across sessions
Every chat, every developer, every agent — same patterns. No more "which style won today?".
Onboard humans AND agents
New devs read the same file Claude reads. One source of truth, zero docs duplication.
Survives team turnover
Your conventions are codified. They outlive any single contributor and scale with the team.
One command. Pick your modules. Firebase or Supabase auto-configured. Start building what matters.
Better AI output starts with better project conventions — pair these rules with the architecture and the Cursor equivalent.
Cursor rules
Same conventions, different IDE. .cursor/rules/*.mdc files pre-configured for ApparenceKit so your Cursor agent matches Claude's output.

3-Layer Architecture
The pattern Claude is trained on. Data, Domain, Presentation — clear boundaries that make every feature predictable.

AI Chat template
Take it further: ship LLM-powered chat inside your app. Streaming responses, any provider, production-ready UI.

Don't take our word for granted, see what some of our apps have achieved
One command. Pick your modules. Firebase or Supabase auto-configured. Start building what matters.
Get Access Now