deer-flow: a super agent harness for orchestrating sub-agents, memory, and sandboxes for deep research

What it solves

DeerFlow is a "super agent harness" designed to orchestrate complex research and task execution. It solves the problem of managing multiple specialized sub-agents, memory, and secure execution environments (sandboxes) to perform deep exploration and efficient research flows that would be too complex for a single LLM prompt.

How it works

DeerFlow acts as a central orchestrator that manages:

• Sub-Agents: It can deploy and coordinate multiple specialized agents to handle different parts of a task.
• Extensible Skills: It uses a set of skills and tools (including Claude Code integration and MCP servers) to interact with the world.
• Sandboxes: It provides isolated environments (Local, Docker, or Kubernetes) to execute code safely.
• Memory & Context: It implements long-term memory and context engineering to maintain state across complex tasks.
• IM Channels: It integrates with messaging platforms like Telegram, Slack, and Discord to receive and execute tasks.

Who it’s for

It is intended for developers and researchers who need a powerful, extensible framework for building autonomous agents capable of deep research, coding, and tool use in isolated environments.

Highlights

• Multi-Agent Orchestration: Ability to coordinate sub-agents for complex workflows.
• Flexible Sandboxing: Supports execution in local environments, Docker containers, or Kubernetes pods.
• MCP Server Support: Extensible via Model Context Protocol (MCP) servers.
• Broad IM Integration: Native support for Telegram, Slack, Discord, Feishu, WeCom, and WeChat.
• Integrated Search: Integration with BytePlus InfoQuest for intelligent search and crawling.