FAQ

Is it free?

Yes. Autonomous Engine is open source under the Apache License 2.0. There is no paid tier, no licence key and no account. You pay only for the infrastructure you run it on and for any hosted LLM you choose to use.

Can I use it commercially or modify it?

The Apache License 2.0 allows both, subject to its terms (keep the licence and notices, state significant changes). Read the LICENSE and NOTICE files. This answer is a summary, not legal advice.

Is there a hosted version?

No. You run it on your own machine or servers with Docker Compose.

Does my code leave my network?

It goes only where you send it. The engine talks to your Git host, your issue tracker if you connect one, and the LLM and embedding providers you configure. With a hosted provider, the code context the agents use is sent to that provider. With a local model through Ollama or vLLM, it stays on your network. The engine has no licence check and does not report usage to the project.

Which models work?

Any model reachable through Ollama, vLLM, OpenAI, Anthropic, or another OpenAI-compatible endpoint, as long as it supports tool calling; the agents work through tools, and the engine stops instead of falling back to plain text generation. Ollama is the default, and a 30B-class coding model such as qwen2.5-coder:32b works. Larger and stronger models generally produce better plans and code. You can give each agent its own model under Admin → Agents.

What hardware do I need?

The stack is about 25 containers. Give Docker at least 16 GB of memory, and more if the model runs on the same machine. A local 30B-class model needs a GPU with enough memory for it.

Will it open pull requests without me?

Only for work that passes verification, and only within the gates you leave on. The PRD needs human approval before work starts, verification must approve before any pull request exists, pull requests can be held at a manual gate, and with HUMAN_REVIEW_REQUIRED=true (the default) a person reviews before merge.

What if my repository has no tests?

By default, work on a repository with no executable checks is blocked at verification. A product can opt in to sending such work to human review instead. Either way, the evidence on the work item shows which checks ran.

Does it work with more than one repository?

Yes. A product can contain several repositories. You record how they depend on each other during onboarding, and the engine opens one pull request per repository it changes.

Which Git hosts and trackers are supported?

The integrations in v0.1.0 are GitHub for source control and Jira for issue tracking. Requests can also come from the dashboard chat, the request portal, or the intake API.

Are there prebuilt images?

The repository’s release workflow publishes per-service images to ghcr.io/autonomous-engine when a version tag is pushed. The docker-compose.yml in the repository builds from source, and that is the path these docs describe.

Can I run it on Kubernetes?

The repository has a Helm chart under deploy/helm. These docs cover only Docker Compose, which is the documented way to run the engine.

How do I get help or report a bug?

Search or open an issue on GitHub. For security problems, follow the security policy instead.

How can I contribute?

See Contributing. In short: open an issue for anything non-trivial, sign off your commits (DCO), and include tests.