· Field NotesMarch 24, 2026

The AI That Actually Does the Work

DeerFlow is an open-source AI agent that does not just answer questions, it plans the work and finishes whole projects on its own.

AIautomationopen-sourceworkflowvia github · @bytedance

The AI That Actually Does the Work

Most AI tools are like a very smart colleague who gives you great advice — then hands everything back to you to actually do.

DeerFlow, released by ByteDance (the company behind TikTok), works differently. You give it a task — say, "research the top competitors in my market and put together a summary report" — and it goes off and does it. It opens files, browses the web, takes notes, and even splits itself into several parallel workers tackling different pieces at the same time. Then it comes back with a finished deliverable.

Think of it like hiring a small research team that works through the night, except it's software running on a server.

What makes this interesting for business owners is the shift in what AI is actually useful for. Right now, most people use AI to get a better first draft or a quick answer. DeerFlow points toward something further: handing off an entire workstream and getting back something close to done.

It's free to use, works with most of the major AI models you've already heard of, and can be set up on your own server so your data stays yours.

Worth watching — especially if you spend chunks of your week on research, reporting, or anything that involves gathering information from many places.

Words worth knowing

AI agent — An AI that takes actions on its own, not just one that answers questions. It can browse, write, and make decisions in sequence.

Open-source — The code is free and public. Anyone can use it, inspect it, or adapt it — no license fees.

Self-hosted — You run the software on your own server instead of someone else's cloud. More control, more privacy.

Model-agnostic — Works with different AI brains (GPT-4, Claude, Gemini, etc.) rather than being locked into one.

Try it

You need Python 3.12+, Node 22+, and Docker. The setup wizard does most of the rest.

  1. Clone the repo: git clone https://github.com/bytedance/deer-flow && cd deer-flow
  2. Run the wizard: make setup — it asks for your LLM provider and writes the config
  3. Pull the sandbox image once: make docker-init
  4. Start it up: make docker-start
  5. Open http://localhost:2026 and hand the agent its first task
Read the install guide
Check it out →

Written by David at AC0.AI. Follow on @ac0hero

Field Notes in your inbox

The AI tools and moves I actually use to win more business. A couple a week, nothing I haven't run myself.

session0
lifetime0
Pressed0