agentic-aiCommunity canon

AutoGPT

The original autonomous AI agent that proved LLMs could act without constant guidance

View on GitHub

What it solves

The 'what if the AI just kept going' question: AutoGPT proved a looping LLM can complete multi-step real-world tasks without human guidance between steps.

Overview

AutoGPT was the experiment that sparked the agentic AI movement, a simple loop where GPT-4 plans, executes tools, reflects on output, and continues without human confirmation at each step. It has since evolved into a structured platform with a GUI, a no-code agent builder, and AutoGPT Bench for reproducible evaluation of autonomous behavior.

Key facts

Language
Python
License
MIT
Maturity
Community canon
Maintainer
Significant-Gravitas
Reviewed
2026-06-11

Where to start

Start with the Classic README to understand the original agent loop before exploring the modern platform, the loop design is the idea worth understanding.

classic/README.md

Technologies

autogptpythonautonomous-agents

More in this domain