Chef, an Open-Source AI App Builder by Convex (github.com)

Convex released Chef, an open-source AI app builder that generates full-stack web apps with a built-in reactive database, zero-config auth, file uploads, real-time UIs, and background workflows. Unlike many LLM app builders that only scaffold front-ends, Chef "knows the backend" by leveraging Convex’s open-source reactive database and APIs—making it particularly well suited for code generation and agentic workflows. The project is a fork of bolt.diy and is available as a hosted webapp (generous free tier) or runnable locally; the system prompt that drives Chef’s agent is openly viewable/downloadable. Technically, Chef wires an agent loop (chef-agent) that injects system prompts, defines tools, and calls model providers (OpenAI, Anthropic, Google, XAI) to produce code and orchestration. The repo includes client code, serverless APIs, the Convex-backed database, a CLI (chefshot), templates, and a test harness for the agent loop. Local development uses Convex’s hosted control plane to provision projects (Chef tokens in local env won’t count against your Convex usage) and requires setting OAuth creds, environment variables, and model API keys. For AI/ML developers this lowers friction for prototyping production-ready, backend-aware LLM apps and demonstrates how a reactive DB plus model orchestration can streamline secure, real-time, full-stack AI application generation.