GitHub radar
DeerFlow: a team of AI agents takes on your task
DeerFlow is a free, open-source tool from ByteDance that runs several AI agents on one task at once — one searches the web, another writes and runs code, a third checks the result. Once installed, you get a local web page where you hand off a task and watch the agents finish it on their own.
It's a conductor for AI agents: not 'smart' itself, it splits work across several agents and gives them shared memory so progress isn't lost between sessions.
When it helps
Useful when a task needs research, coding, and result-checking at once without your constant input. Skip it for a simple task one agent can handle, or if you have no paid LLM API keys.
Pros
- Multiple agents work in parallel, not one after another
- Shared memory across sessions, so progress isn't lost
- Works with OpenAI, Anthropic, Gemini, DeepSeek, and other providers
- 82,000+ GitHub stars and #1 spot on GitHub Trending
Cons
- Needs paid API keys from an LLM provider, there's no free intelligence here
- On its own it doesn't replace a full coding agent like Claude Code
How to set it up — step by step
- 1Open your AI agent (Claude Code, Codex) in a separate project folder.
- 2Send the agent the text from the block below — it clones the repo and sets up Docker.
- 3Tell the agent which LLM provider and API keys you want to use.
- 4Wait while the agent starts the containers and opens the web page at localhost in your browser.
- 5Check the interface: give it a test task and watch the agents carry it out.
- 6For regular use, make sure you have Docker and at least 16 GB of RAM.
Text for your agent
Copy this and send it to your agent — Claude Code, Codex, any of them:
Help me set up DeerFlow from here: https://github.com/bytedance/deer-flow — clone the repo, follow Install.md, use Docker, ask me which LLM provider and API keys I have, and open the web interface on localhost. Ask if you need clarification.
Requires Docker and at least 8 GB of RAM — an M-chip MacBook handles it fine for testing. For production use, 16 GB or more is recommended.
Open on GitHub▌ More finds