AutoGPT Platform: release 0.10.0 with self-hosted Docker Compose and a new API
On October 27, 2026, release 0.10.0 of the AutoGPT Platform came out — a visual AI agent builder where specialists are assembled in the interface and coordinated through the Otto coordinator. Since the last review, 0.8.1—0.8.4 and 0.9.0 have also been released, so this digest collects the changes from six releases.
What's new
- Self-hosted Docker Compose is now secure by default — passwords for PostgreSQL, RabbitMQ, Redis, FalkorDB and ClamAV are now generated at install time, and services listen only on 127.0.0.1. Before launching, you need to run make init-env or the installer.
- External API v1 is marked deprecated — v2 is being introduced in its place, and the old one will keep working until January 31, 2027.
- Self-hosting from a Docker Hub image with a Windows installer is now available — one container, one command, then localhost:3000 and creating an administrator.
- Every block now declares its effect — Auto itself evaluates cheap external actions and asks permission for expensive ones, which reduces the number of unexpected operations.
- AutoPilot pauses tasks at limits instead of stopping — the Keep going mode picks them up when the budget resets, and scheduled chats report why they skipped a run.
- Specialist integrations connect via MCP first — this simplifies linking external tools to agents.
What the platform does
AutoGPT Platform is a visual builder where agents are assembled with a mouse: specialists use skills and workflows, and the visual editor lets you define repeatable steps, connect tools and monitor execution. The platform is deployed on your own infrastructure via Docker, and the model is connected through Ollama.
Your own model instead of the cloud
Your own model is connected through Ollama — the documentation describes the "Running Ollama with AutoGPT" scenario. Self-hosting is set up with a single docker run -d command from the significantgravitas/autogpt image on Docker Hub, or via the installer on Linux, macOS and Windows. License — PolyForm Shield License 1.0.0: commercial use is permitted, except for creating competing products; when distributing, you must include the license text and the Required Notice lines.
How it differs from its neighbors
Compared with Dify, Langflow and n8n, AutoGPT Platform has not a single feature that is unique to it. MCP support, a knowledge base (RAG) and tracing with debugging are claimed by most of its neighbors, but are not mentioned in the AutoGPT Platform documentation. Stateful graphs, a visual editor and your own server via Docker are present in all four platforms.
| AutoGPT Platform | Dify | Langflow | n8n | |
|---|---|---|---|---|
| Type | builder | builder | builder | builder |
| Language | Python | Python | Python | TypeScript |
| License | PolyForm Shield Licen… | Apache 2.0 with add. cond… | MIT | Sustainable Use Licen… |
| GitHub stars | 187,826 | 159,392 | 155,815 | 206,788 |
| Releases in 90 days | 12 | 4 | 15 | 58 |
| Multiple agents in a chain | — | — | yes | — |
| Custom tools and functions | yes | yes | yes | — |
| MCP support | — | yes | yes | — |
| Knowledge base (RAG) | — | yes | yes | — |
| Sandboxed code execution | — | yes | — | — |
| Human approval | — | — | — | yes |
| Graphs and state persistence | yes | yes | yes | yes |
| Visual editor | yes | yes | yes | yes |
| Tracing and debugging | — | yes | yes | yes |
| Voice and speech | — | yes | — | — |
| Own server and Docker | yes | yes | yes | yes |
| Own model: Ollama | yes | — | yes | — |
| Own model: LM Studio | — | — | — | yes |
| Own model: OpenAI-compat. API | — | yes | — | yes |
Comparison based on the projects' READMEs and documentation as of 10/30/2026: "yes" means the feature is stated there, a dash means it is not mentioned (which does not mean the feature is absent). Stars and releases are per GitHub data.
Those already self-hosting AutoGPT Platform should update: release 0.10.0 changes the Docker Compose startup order, and without make init-env the services won't come up. Those looking for a visual agent builder with a coordinator and local model connection via Ollama should take a closer look. You can follow new releases in the CPU3D news feed.
