Convex Agents

This handles the full lifecycle of AI agents on Convex, from thread management to tool calling to RAG. You get persistent conversation state that survives restarts, streaming responses that update clients in real time, and the ability to define tools as Convex functions the agent can call. The RAG setup uses Convex's vector search for knowledge retrieval, and everything runs on durable workflows so long-running tasks don't break. Good fit if you're building chatbots or assistants that need to remember context, search knowledge bases, or trigger actions in your app. The streaming implementation is cleaner than rolling your own state management.