A cross-platform AI chat app that remembers you.

The AI chat apps I used kept forgetting things I’d told them ten minutes earlier. Every conversation felt like an introduction. I wanted an app that built up a quiet model of who I am and what I’m working on across sessions, so I could pick up a thread three weeks later without re-explaining myself.

And I wanted it to run everywhere. Phone on the bus, tablet at the kitchen table, browser at the desk — same threads, same memory, no copy-paste between platforms.

• Cross-platform from day one. React Native first, with the same codebase compiling to iOS, Android, and a web build. No platform-specific forks.
• Automatic context memory. A retrieval layer summarises and indexes prior conversations on the fly. The app remembers facts you’ve mentioned, projects you’re shipping, and people you’ve named — without you ever managing a prompt.
• Streaming responses. Token-by-token rendering with cancel-mid-stream, retry, and message editing.
• Tool calling. The model can call a small set of internal tools (memory search, link preview, calculator) when the conversation needs it.
• Account sync. Threads, memory, and settings sync across devices the moment you sign in.

• Frontend. React Native, TypeScript, Expo router. Custom design system tuned for long reading sessions on phone and tablet.
• Backend. Hono on the edge runtime — sub-100ms cold starts, no server I have to babysit. TypeScript end-to-end so types flow from DB through API to UI.
• Data. Drizzle ORM over Postgres, with a memory index in the same database. One migration story, one source of truth.
• AI. Provider-agnostic chat layer with a streaming inference path and the memory-retrieval tool wired in as a callable.

Mooncake is the proof that the same engineer who ships regulated platforms can move fast on consumer mobile. The discipline travels: the audit-style logging that keeps a credit union out of trouble is the same discipline that turns an AI chat app into a reliable daily tool.

It also unlocks the React Native + edge stack for client work. When a customer asks for a cross-platform app with a real backend, this is the codebase the pattern came from.