DeerFlow is ByteDance's open-source super-agent framework for research, coding, and multi-step knowledge work. The current 2.0 generation is built as a general agent harness rather than only a deep-research workflow, combining tools, skills, sub-agents, memory, scheduling, context management, and sandboxed execution so an agent can plan and complete complex tasks across multiple steps.

The project supports configurable model providers, web search, file and shell tools, MCP servers, messaging channels, tracing, long-term memory, and isolated execution environments. It also includes a terminal workbench and scheduled tasks, making DeerFlow useful for developers who want to build or self-host capable autonomous assistants that can research, write files, execute code, delegate work, and maintain context over longer-running workflows.