deer-flow
View on GitHubAn open-source long-horizon SuperAgent harness that researches, codes, and creates. With the help of sandboxes, memories, tools, skill, subagents and message gateway, it handles different levels of tasks that could take minutes to hours.
ByteDance's DeerFlow 2.0 is a SuperAgent harness that orchestrates subagents, sandboxes, memory, skills and tools to run research, coding and content tasks lasting minutes to hours. Ships with a web UI, TUI, IM gateway, MCP server and LangGraph backend.
Use Cases
Long-horizon research and report writingMulti-agent task delegation to subagentsAutonomous coding inside a sandboxDeep web research with search and crawlingContent creation such as podcasts and slidesScheduled recurring agent tasksLong-term memory backed personal assistantConnecting agents to IM channels as a chatbotExposing agent capabilities via MCP serverTerminal-based agent workbench (TUI)
Built With
- Language
- Python
- Frameworks
- LangChain · LangGraph · FastAPI · Next.js · Node.js · Docker · Playwright · vLLM · MCP · LangSmith · Langfuse · pnpm
Tags
multi-agent · superagent · deep-research · subagents · sandbox · long-term-memory · skills · langgraph · harness · workflow-orchestration · coding-agent · podcast-generation · report-generation · message-gateway · human-in-the-loop · self-hosted