convex-agents

To build persistent AI agents, you use the Convex Agent component to manage thread state, tool execution, and workflow orchestration. This ensures conversation history survives restarts and streams real-time updates to clients automatically.

Yes, you can execute Convex functions as agent tools for AI workflows. The Convex Agent component enables tool execution by exposing server APIs, allowing your agent to trigger functions and orchestrate long-running durable tasks reliably.

The best way to implement retrieval-augmented generation is using the Convex Agent component's built-in RAG capabilities. It handles vector search for knowledge retrieval directly, enabling persistent agents to query stored context and ground responses in real-time.

You stream real-time AI responses by configuring the Convex Agent component with client APIs. This setup automatically pushes streaming updates to connected clients, enabling live conversation feeds during persistent, long-running agent workflows.

Yes, the Convex Agent component is required to manage durable workflows for long-running agent tasks. It provides the server and client APIs needed to orchestrate persistent state, tool execution, and reliable execution across sessions.

AI agents lose conversation history without persistent state management. By applying the Convex Agent component, thread state and conversation history survive server restarts, ensuring continuous and stateful interactions across sessions.