LlamaCoder
An open-source AI app generator you can self-host, with browser previews and cloud inference through Together AI. Licensed under MIT.
LlamaCoder generates small apps from a single prompt and previews them in your browser. It's an open-source alternative to Claude Artifacts for people who want to turn an idea into an app, with a MIT license that lets them run and modify the code themselves.
You can use the hosted website or run your own copy locally. That distinction matters for privacy and hardware: the app can run on your machine, but Together AI handles model inference in the cloud. Prompts go to that service, so a local deployment still needs an internet connection and a Together AI API key. It isn't an offline local LLM runner.
The hosted interface offers DeepSeek V4 Flash. The repository also describes generation with Meta's Llama 3.1 405B through Together AI. App previews run in a sandboxed browser iframe, so you can see the generated result alongside the code generation experience.
For developers choosing a project to adapt, the codebase uses TypeScript, Next.js and Tailwind. Its preview renderer uses esbuild-wasm and esm.sh to render generated apps in the browser. A self-hosted copy also requires a PostgreSQL database connection, and screenshot uploads use S3 storage.
