Visual builder + your own coding agent: does this fit how you build?

Hi Product Hunt 👋

We’re the team behind Mapples Creator, a visual builder for Expo and React Native apps. You design screens on a canvas, then wire up data, actions and navigation. What you get out is real Expo code, not a locked-in runtime.

You can build with AI in three ways:

1. Build in web (Beta): an assistant works on your canvas. You pick a style direction, review its plan, and it builds screen by screen. You can undo any turn in one click.

2. Build in agent: Claude Code, Cursor or Codex writes the code on your machine, at your own token cost. Sync goes both ways: code edits show up on the canvas, and canvas edits land in your TSX.
   npx mapples create my-app --secret <key> --agent claude

3. MCP connector: use your Mapples projects from Claude or any MCP client, with no keys to copy.
   claude mcp add mapples --transport http https://prod-us.mapples.org/mcp/

The runtime is open source (@mapples/*, MIT), and the output is plain Expo, Expo Router and EAS.

We’d love your input:

• If you build mobile apps with AI today, where does it fall apart? Layout, navigation, state, or keeping code and design in sync?

• Would you rather have the assistant inside the editor, or your own agent working with a canvas next to it?

• What would make you trust exported code enough to ship it?

• Designers: what’s missing for you to hand a canvas straight to a developer or an agent?

Bug reports, feature requests and blunt feedback are all welcome. We read every reply.

Try it: app.mapples.org · Docs: docs.mapples.org