ByteDance · 2026-03-10 · notable
DeerFlow 2.0 — open-source SuperAgent harness
ByteDance's open-source SuperAgent framework: orchestrate sub-agents with memory, sandboxes, and skills for long-running research, coding, and creative tasks. 60.6k stars, MIT license.
ByteDance's MIT-licensed SuperAgent harness that orchestrates sub-agents, sandboxes, and persistent memory for tasks that take minutes to hours.
Key specs
| License | MIT |
|---|---|
| GitHub stars | 60.6k |
What is it?
DeerFlow 2.0 (Deep Exploration and Efficient Research Flow) is a ground-up rewrite of ByteDance's open-source agent framework, released in early March 2026. It is a SuperAgent harness: a lead agent can spawn sub-agents on the fly, each with scoped context, tools, and termination conditions, running in parallel when possible. It hit #1 on GitHub Trending in late February and now has over 60k stars.
How does it work?
Built on LangGraph and LangChain, DeerFlow ships with a filesystem, long-term memory, an extensible skills framework, and sandbox-aware execution (local, Docker, or Kubernetes). Sub-agents run in parallel, report back structured results, and the lead agent synthesizes everything. It supports GPT, Claude, Gemini, DeepSeek, and local LLMs, and integrates with Telegram, Slack, Feishu, WeChat, and other IM channels.
Why does it matter?
The gap between 'demo agent' and 'production agent' is sandboxing, memory, error recovery, and multi-agent coordination. DeerFlow 2.0 ships all of that under MIT in a single framework. ByteDance using it internally gives it credibility that hobby projects lack, and the 60k stars show real community traction.
Who is it for?
Teams building multi-step agent workflows, developers who need a production-grade agent harness.
Try it
github.com/bytedance/deer-flow