What it does AutoGPT is a platform for building, deploying, and running continuous AI agents. You assemble workflows by connecting single-action blocks in a low-code frontend; the server runs them, triggers them from external sources, and keeps them humming. There’s also a marketplace of pre-built agents for the click-and-run crowd.

The interesting bit The project has split its own history in two: the new autogpt_platform (Polyform Shield license) is the commercial future, while the original agent, Forge toolkit, benchmark, and UI remain MIT-licensed “classic” artifacts. It’s a tidy separation of church and state — or perhaps of SaaS and community goodwill.

Key highlights

• One-line install script for self-hosting via Docker (curl or PowerShell)
• Block-based agent builder with deployment controls and monitoring
• Pre-built examples: Reddit-to-viral-video pipeline, YouTube-to-social-quote extractor
• Classic tooling still ships: Forge boilerplate kit, agbenchmark for agent evaluation, protocol-compatible frontend
• Uses the AI Engineer Foundation’s agent protocol for interoperability

Caveats

• Self-hosting requires 8–16GB RAM, Docker, Node, and tolerance for “a technical process” (their words)
• Cloud-hosted beta is waitlist-only; the free self-host route is currently the only guaranteed access
• The Polyform Shield license on the platform code is not OSI-approved; commercial use may need scrutiny

Verdict Worth a look if you want to prototype agent workflows without wiring LLM calls by hand. Skip it if you need a polished, fully-managed product today — or if license ambiguity makes you nervous.