AutoGPT

What this means for you: AutoGPT Platform v0.10.0 changes how the older way of connecting other software to AutoGPT (external API v1) works and how a self-hosted install starts. The new API v2 replaces v1. Each API key now records when it was last used, so you can spot keys nobody uses anymore.

Why it matters: The change to self-hosted installs continues a pattern. On Sep 19, v0.8.0 already required self-hosted installs to generate their own encryption and secret keys instead of using shipped defaults.

How to use it: If you connect to AutoGPT through external API v1, plan to move to v2. v1 keeps working until 2027-01-31. If you self-host with Docker Compose, run make init-env (or the installer) before docker compose up. The install won't start without its generated service passwords. Existing installs should follow the Upgrading notes on service passwords.

What this means for you: AutoGPT Platform lets you build and run AI agents that do tasks for you. Version 0.9.0 lets you host it yourself from a Docker Hub image, with a Windows installer. In Auto mode, the agent now judges cheap outward actions itself and asks you first before costly ones. Administrators also get a users page to look up, ban and impersonate a user.

Why it matters: It builds on the Sep 30 release that added Ask First, Auto and Unsupervised modes and paused workflows before irreversible actions. The Oct 14 release redesigned the wallet, and this one redesigns usage limits.

How to use it: Self-host from the Docker Hub image, using the Windows installer if you are on Windows. Usage limits have been redesigned, and the release adds a changelog notification for new releases. This is autogpt-platform-beta-v0.9.0.

What this means for you: The wallet in AutoGPT Platform is redesigned to show chat usage along with credits. The account menu now has a Give feedback option. Expert integrations now connect through MCP first, which is a standard way for AI tools to plug into other apps.