DeerFlow

Open-source SuperAgent harness by ByteDance for deep research, coding, and creation.

Overview

DeerFlow is an open-source SuperAgent harness that researches, codes, and creates. With sandboxes, memories, tools, skills, and subagents, it handles tasks ranging from minutes to hours. Originally a deep research agent, it has evolved into a full-stack Super Agent.

Key Features (DeerFlow 2.0)

Context Engineering: Long/short-term memory for better understanding
Long Task Running: Planning and sub-tasking — plans ahead, reasons through complexity, executes sequentially or in parallel
Extensible: Skills and tools system — plug, play, or swap built-in tools
Persistent: Sandbox with file system — read, write, run like a real computer
Flexible: Multi-model support (Doubao, DeepSeek, OpenAI, Gemini, etc.)
Free: MIT License, self-hosted, full control

Agent Skills

Skills are loaded progressively — only what's needed, when it's needed. Built-in library includes:

Deep search (biotech, computer science, physics)
Frontend design
Deploy (with scripts)

Agent Runtime Environment

Docker-based secure sandbox that can execute commands, manage files, and run long tasks. Recommends All-in-One Sandbox combining Browser, Shell, File, MCP, and VSCode Server in a single container.

Features: Isolated, Safe, Persistent, Mountable FS, Long-running

Relation to OpenClaw

DeerFlow and openclaw share architectural similarities:

Both use Skills mechanism for extensibility
Both feature sandboxes, memory systems, and tool integration
Both are MIT-licensed, self-hosted agent frameworks
Focus difference: DeerFlow emphasizes research and coding tasks; OpenClaw is a personal assistant platform that integrates with messaging apps
Deployment: DeerFlow runs task-oriented sessions; OpenClaw runs 24/7 as persistent assistant