DeepAgents: The Claude Code Alternative That Works With Any LLM

LangChain dropped DeepAgents last week, and it's already my favorite way to spin up a capable agent without wiring prompts together like it's 2024.

Think of it as Claude Code's architecture, decoupled from Claude. Planning, filesystem access, sub-agents, context management — all available as a pip-installable harness that works with any tool-calling LLM.

Here's everything you need to know.

What You Get Out of the Box

pip install deepagents

That's it. You now have an agent that can:

Plan — write_todos breaks down tasks and tracks progress
Work with files — read_file, write_file, edit_file, ls, glob, grep
Run commands — execute with sandboxing
Spawn sub-agents — task delegates work with isolated context windows
Manage context — Auto-summarization when conversations get long

The philosophy is simple: trust the LLM, enforce boundaries at the tool level. Don't expect the model to self-police — sandbox what it can touch.

30-Second Quickstart

from deepagents import create_deep_agent

agent = create_deep_agent()
result = agent.invoke({
    "messages": [{"role": "user", "content": "Research LangGraph and write a summary"}]
})

The agent plans, reads files, and manages its own context. No prompt engineering required.

Swap Models, Add Tools, Customize

DeepAgents is just a compiled LangGraph graph. Customize it like any other:

from langchain.chat_models import init_chat_model

agent = create_deep_agent(
    model=init_chat_model("openai:gpt-4o"),
    tools=[my_custom_tool],
    system_prompt="You are a research assistant.",
)

MCP support? langchain-mcp-adapters has you covered.

The CLI (Optional but Nice)

curl -LsSf https://raw.githubusercontent.com/langchain-ai/deepagents/main/libs/cli/scripts/install.sh | bash

Gets you web search, remote sandboxes, persistent memory, and human-in-the-loop approval flows.

When to Reach for It

Use Case	DeepAgents	Claude Code
Quick research task	✅	✅
Long-running workflow	✅	✅
Custom model (local/enterprise)	✅	❌
Built-in IDE integration	❌	✅
Need full control over prompts	✅	⚠️

If you're building agents into your product, DeepAgents is the move. If you want the best terminal experience for yourself, Claude Code still wins.

The Bottom Line

LangChain extracted the agent harness pattern from Claude Code and made it model-agnostic, open-source, and production-ready.

It's not trying to replace Claude Code. It's trying to give you Claude Code's architecture for your own agents.

And for that, it nails it.

Next up: Try swapping in a local model with Ollama and see how the same agent behaves with different brains. The results might surprise you.