Open-source, MIT-licensed super-agent that researches, codes and creates using sandboxes, memory, skills and subagents; self-hosted.
What it does
DeerFlow is an open-source agent framework that began as a deep-research tool and grew into a full-stack 'super agent'. It combines a Docker-based sandbox, long/short-term memory, extensible skills and tools, and subagents to handle tasks ranging from minutes to hours, with support for multiple model providers.
Core features
Deep-research agent with multi-step planning
Docker-based all-in-one sandbox (browser, shell, files, VS Code)
Long and short-term memory / context engineering
Extensible, swappable skills and tools
Subagents and parallel task execution
Multi-model support (DeepSeek, OpenAI, Gemini, Doubao, etc.)
Best for
→Automated deep-research reports
→Generating content, comics or videos from prompts
→Exploratory data analysis
→Running long, multi-stage agent tasks
Reviews
Big-picture takes: what it's for and whether it delivers. High-engagement YouTube videos — not sponsored.