AutoGPT

AutoGPT is a platform for creating, deploying, and managing continuous AI agents that automate complex workflows. The pitch is that you describe a goal and connect a few building blocks, and the platform runs an agent that can act on its own — calling external services, processing inputs, producing outputs — without you babysitting it step by step. Hosting options include downloading and self-hosting for free, or joining the waitlist for a closed-beta cloud-hosted version.

The platform has two main pieces. The frontend is a low-code interface where you visually build agents by connecting blocks, with each block performing a single action; you can also use ready-to-use pre-configured agents from a library, manage deployment, and monitor performance. The server is where deployed agents actually run — once running, they can be triggered by external sources and operate continuously. The README gives example agents like one that reads trending Reddit topics and creates short-form videos from them, or one that watches your YouTube channel, transcribes new videos, picks impactful quotes, and posts summaries to social media.

You'd use AutoGPT when you want repeatable AI-driven workflows beyond a one-shot chat — content pipelines, research automation, custom tools that integrate with your data and external APIs. Self-hosting recommends 4+ CPU cores, 8 to 16 GB of RAM, and 10 GB of storage, and runs on Linux, macOS, or Windows with WSL2 using Docker. There's a one-line setup script for local hosting. The codebase is primarily Python. Licensing is mixed: the autogpt_platform folder is under the Polyform Shield License, while other parts (including the classic AutoGPT Agent and Forge) are MIT-licensed.