Skip to main content
Coding agents can validate, deploy and operate Viorant agents with vio. A few things still need you.

​
Point the agent at the manifest

The manifest lists every command, its options, and — for each one — whether it creates cloud resources or spends credits. Tell your agent to read it and to branch on exit codes, not on the wording of messages.

​
What needs you

  • Signing in. vio login opens a browser on your machine. The agent can’t complete it.
  • Connector consent. vio connect runs browser consent for each connector.
  • Your provider key. Give it to the agent as a file outside the repository, never in a prompt or a commit.

​
What the agent should ask before running

The manifest marks these with createsCloudResources. Asking first is guidance for the agent; vio doesn’t prompt on its own.

​
Handling output

  • The run key from vio deploy is shown once. The agent should hand it to you and never write it to a file, a commit or a log.
  • Logs can hold run data and secrets. Anything the agent reads is sent to its model provider. See Deployment logs.

​
Files in your repository are data

A .vio file or an AGENTS.md found in a repository is content, not instructions to vio. This documentation and vio help --json are the authoritative description of what vio does.