Auto Runtime

Describe how you want the work done. Auto builds a working playbook you can test, change and run again. You decide when it runs and what it is allowed to do.

A playbook contains one brain, its reusable function collection, and workflows that coordinate those functions. Functions can also run directly. Each function or workflow call produces a run with its result and the versions used. Connections provide authorized access to external services.

Your agent builds and directs the work. Auto Runtime stores and executes it; Auto Studio lets you inspect and manage it.

Get started

Start with hosted Auto: connect your agent to https://mcp.on.auto, sign in when it asks, and see your work in Auto Studio. The quick start needs nothing installed and no model key.

To run Auto Runtime on your own infrastructure, every release is the image docker.on.auto/onauto/runtime, from Docker Hub. Your team reaches it at its PUBLIC_URL, the https address people sign in at, and opens it in Studio from https://studio.on.auto; see Self-hosting, Run with Docker and Deploy for your team. Auto sends an anonymous usage ping when it starts, which DO_NOT_TRACK=1 turns off; Usage ping lists exactly what it sends.

To run it for yourself on your own computer, npx @on.auto/runtime starts it in local mode, with no sign-in, on Node.js 26.10 or later on macOS or Linux; see Run Auto on your computer.

Browse the Auto documentation, and see function availability for implemented capabilities.

Auto Studio with synthetic Accounts Payable data.

Contribute and get help

See CONTRIBUTING.md for development setup and checks, and documentation contributions to preview a guide. Open an issue for bugs and questions; report vulnerabilities through SECURITY.md.

License

Auto Runtime is source-available under the Elastic License 2.0. LICENSING.md explains the terms. Contributions follow the CLA and Code of Conduct.