AWS's React Native AI app that generates other apps
A mobile chat client that can spin up working web apps, images, and voice conversations from a single prompt.

What it does
SwiftChat is a React Native AI assistant for Android, iOS, and macOS. It streams chat responses, generates images, runs web search, and—unusually—can create, edit, and share instant web apps (think 2048, Tetris, news readers) from natural language prompts. It supports multiple model backends: Amazon Bedrock, Ollama, DeepSeek, OpenAI, and generic OpenAI-compatible providers.
The interesting bit
The “Create App” feature isn’t a gimmick: the app generates runnable web apps in the background with iOS Live Activity or Android notification progress tracking, then lets you edit and share them. There’s also a serverless deployment option via API Gateway + Lambda that enables 15-minute streaming responses—useful for long model runs that would normally hit timeout walls.
Key highlights
- Multi-backend by design: Bedrock API key, self-hosted SwiftChat Server, Ollama local, DeepSeek, OpenAI, or OpenAI-compatible endpoints
- App-in-an-app: Generate, preview, edit, and share web apps without leaving the chat interface
- Nova Canvas integration: Virtual try-on, background removal, and image gallery with automatic clothing recognition
- Speech-to-speech: Amazon Nova Sonic with barge-in, echo cancellation, and voice waveform visualization (server-only)
- One-command server deploy:
curl | bashinstall script spins up CloudFormation stack with QR code auto-configuration
Caveats
- iOS requires local Xcode build; no TestFlight or App Store link provided
- Amazon Nova Sonic speech-to-speech only works with the SwiftChat Server backend, not direct Bedrock API key
- Web search beta sources (Baidu, Bing) noted as potentially inaccurate; Tavily recommended
Verdict
Worth a look if you want a single mobile client that can switch between cloud and local models, or if the “generate a working mini-app from chat” workflow fits your use case. Skip it if you need a polished iOS store experience or want to avoid AWS infrastructure entirely.