deepagents-langchain

You orchestrate multi-step AI tasks by using the DeepAgents framework to create LangGraph agents that handle task decomposition, isolated sub-agents, and auto context summarization for complex workflows.

The best way to add planning is using the write_todos feature in the DeepAgents framework, which provides built-in task decomposition and progress tracking for LangGraph agents.

Yes, you can delegate tasks to sub-agents with isolated context windows by calling create_deep_agent, which returns a fully functional LangGraph agent with built-in sub-agent delegation capabilities.

Yes, the DeepAgents framework supports sandboxed shell execution and filesystem operations including reading, writing, editing, and searching files directly within your LangGraph agent workflows.

You persist context and checkpoint state by utilizing the auto-summarization and persistence features built into the DeepAgents LangGraph runtime, ensuring long-running workflows maintain state.

No, MCP adapters are optional, but they provide a seamless way to integrate external tools and systems into your DeepAgents LangGraph runtime without manual setup.