ByteDance Deer-Flow: Open-Source SuperAgent Harness

DeerFlow (Deep Exploration and Efficient Research Flow) is ByteDance’s open-source “super agent harness” built on LangGraph and LangChain that evolved from a deep research framework into a general-purpose agent runtime. Version 2.0 ships with sandboxed Docker execution environments (with optional Kubernetes support), a persistent filesystem, long-term memory, progressive skill loading, sub-agent orchestration for parallel multi-step tasks, and context engineering including summarization and isolated sub-agent contexts. It supports extensible skills (research, report generation, slides, web pages, image/video generation) and tools via MCP servers, integrates with messaging platforms (Telegram, Slack, Feishu), offers a Claude Code integration skill, and can be used as an embedded Python client library.

Source: bytedance/deer-flow