deepagents

Why we included this project

Deep Agents is a harness that gets you a working agent on day one, which matters for long, multi-step jobs that touch files. The parts that usually take the most effort are already bundled: sub-agents with isolated context, filesystem read/write/search, shell access, context summarization, and persistent memory, all behind a single create_deep_agent call. Because it is built on LangGraph, streaming, checkpointing, and persistence come for free, and you can plug in any tool-calling model, from frontier APIs to local Ollama or vLLM deployments. Every component can be overridden or replaced without forking, and human-in-the-loop approval of tool calls is handled out of the box. It is a practical starting point for production agent work rather than a research prototype.