From a 2023 experiment to a platform
AutoGPT started in March 2023 as a standalone script that chained GPT calls into an "autonomous" loop, and it became one of the most-starred projects on GitHub almost overnight. What sits behind those 185,000 stars today is a different thing: a platform for building, deploying, and managing continuous AI agents that automate workflows, with a low-code builder, a server that hosts the agents, and a marketplace of pre-built ones. The original standalone agent still lives in the repository under classic/, but the active work is the platform.
How you actually run it
There are two paths. You can self-host the AutoGPT platform for free, or join the waitlist for a cloud-hosted beta that is still in closed release. Self-hosting is a real setup, not a one-file download. The README lists Docker Engine and Docker Compose, Git, Node.js and npm, and a code editor, on a machine with at least four CPU cores and 8 to 16GB of RAM. There is a one-line install script for macOS, Linux, and Windows (through WSL2) that wires up dependencies and Docker for a local instance. If that stack sounds like more than you want to operate, the project itself points you to the cloud waitlist instead.
