AI agents

AutoGPT

The project that kicked off the AI-agent craze: give it a goal and it plans and runs the steps itself.

Rank

#1

Stars

188k

Licence

Mixed (see LICENSE)

Checked

6 Nov 2026

What it does

AutoGPT connects a language model to tools such as web search and files, then loops: it plans a step, runs it, reads the result and plans the next one until the goal is met. It went viral in April 2023 and has since grown into the AutoGPT Platform, where you build agents from blocks in a visual editor and run them on your own server.

How to use it

Shell: PowerShell (Windows 11, Docker Desktop and Git installed)

1. 1

   Clone the repository.

   git clone https://github.com/Significant-Gravitas/AutoGPT.git

2. 2

   Open the platform folder and follow the self-hosting guide in its README to set your environment file.

   cd AutoGPT/autogpt_platform

3. 3

   Start the platform in Docker, then open the address the guide gives you in your browser.

   docker compose up -d

4. 4

   Stop everything when you are done.

   docker compose down

Best practices

• Set a monthly spending limit on the API key you give it. Agents call the model many times per task and costs add up fast.
• Keep a human approval step for anything that sends, buys or deletes. Agents can loop on a wrong path or make things up.
• Give small, specific goals. "Summarise these 3 pages" works; "grow my business" doesn't.
• Check the licence. The newer platform code is not under the same licence as the classic version.

In the news

• TechCrunch · 22 Apr 2023 · What is Auto-GPT and why does it matter?
• Tom's Hardware · 22 Apr 2023 · Auto-GPT and BabyAGI Are AI's New Hotness, But They Suck Right Now