deepagents is an opinionated, model-agnostic agent harness bundled with filesystem, sub-agents, context management, and skills for long-horizon, multi-step work. It installs via a single uv command and works with any LLM that supports tool calling.
Project overview
It provides a more opinionated harness on top of create_agent with filesystem, sub-agents, context management, and skills bundled in, addressing the lack of a batteries-included agent harness that runs for multi-step work.
Project type
MCP · AI Agent
Use cases
Automation
Deployment
Refer to project documentation
License
MIT
Best for
Developers and AI engineers who need an opinionated agent harness with bundled filesystem, sub-agents, context management, and skills for long-horizon, multi-step work.
Teams that require model-agnostic support and want to use frontier APIs, open-weight, or local models interchangeably.
Key capabilities
Delegate tasks to agents with isolated context windows.
Read, write, edit, or search over pluggable local, sandboxed, or remote backends.
Summarize long threads and offload tool outputs to disk.
Run commands in your sandbox of choice.
Pluggable state and store backends for cross-session recall.
Approve, edit, or reject tool calls before they run.
Reusable behaviors the agent can load on demand.
Bring your own functions or any MCP server.
Limitations and risks
The agent follows a trust-the-LLM model where it can do anything its tools allow.
Users must enforce boundaries at the tool and sandbox level, not by expecting the model to self-police.
Getting started
Install the package with a single uv command (uv add deepagents), instantiate create_deep_agent in Python, then invoke the agent.
Evidence and sources
README: Deep Agents is an open source agent harness — an opinionated agent that runs out of the box. Extend, override, or replace any piece.
README: Deep Agents is a more opinionated harness on top of `create_agent` — same building blocks, but with filesystem, sub-agents, context management, and skills bundled in.
README: Yes. Any model that supports tool calling works — frontier APIs (OpenAI, Anthropic, Google), open-weight models hosted on providers like Baseten or Fireworks, and self-hosted mode…
README: Deep Agents is built on LangGraph, designed for production agent deployments. Pair it with [LangSmith](https://docs.langchain.com/langsmith/home) for tracing, evaluation, and moni…