VIORANT CLI # DEPLOY AN AGENT. FROM YOUR TERMINAL. vio is the Viorant CLI. It turns a folder into a [.vio](/vio) Space, deploys it to [Viorant Cloud](/cloud), and hands you a URL. Run it yourself, or let Claude Code, Codex or Cursor drive it. The same deploy pipeline as the [Hub](/hub), without opening the Hub. [Install vio](#install)[Documentation](https://docs.viorant.ai/cli/install) $ vio login $ vio init $ vio deploy agent.vio --target viorant_cloud ## INSTALL Run it with npx, nothing to install: npx -y @viorant/cli --version Or install it with npm i -g @viorant/cli, or on macOS and Linux with the signed installer, which verifies the release before it writes anything: curl -fsSL https://dl.viorant.ai/cli/install.sh | sh vio needs Node 22 or 24. The installer puts it in your home directory without sudo. Run the same command again to upgrade. Windows: download the release file from dl.viorant.ai and check its sha256 checksum by hand. ## THREE COMMANDS TO A RUNNING AGENT - ### 01 Sign in vio login Google sign-in in your browser. One owner-only credential file on your machine. - ### 02 Initialize vio init Turns the folder into a Viorant Space: agent.vio and its files, ready to edit. Works offline, no account needed. - ### 03 Deploy vio deploy agent.vio --target viorant_cloud You get a URL and a run key, shown once. Runs on your Viorant credits, no provider key required. Between init and deploy, edit agent.vio yourself, or let your coding agent build it: npx -y @viorant/cli agents install teaches Claude Code, Codex or Cursor how. If the agent uses connectors, sign in to them first with vio connect github slack, and the deploy brings the agent up connected. Then call it with vio run --task "...", or over HTTPS with its run key: curl -sS -X POST https://.a.run.app/run -H "Authorization: Bearer " -H "Content-Type: application/json" -d '{"task":"..."}' To give the agent typed decisions mid-run, add the [Jev MCP server](/jev): vio docs agents prints the recipe. ## AFTER DEPLOY - vio status : your plan, credit balance, and how many agents you can still deploy. - vio list and vio show : your deployments and their state. - vio run --task "..." : call the agent once, with the run key vio saved at deploy. - vio runs : recent runs of a deployed agent, each with a short run id. - vio logs : recent logs of a deployed agent, or one run with --run , with terminal control stripped and common secrets redacted. - vio redeploy : replace a running agent in place. The URL stays the same. - vio connect : finish or re-authorise a deployed agent's connectors in your browser. No redeploy. - vio stop : stop the agent and revoke its credentials. - vio schedule --every 1d --at 09:00 : run the agent on a repeating schedule, in Viorant Cloud. ## WHAT TEAMS USE IT FOR ### Ship from CI vio validate --json gates a pull request. A merge redeploys the agent in place. ### Keep agents in git A .vio file is reviewable text with a canonical digest. Diff it, review it, and deploy exactly what was merged. ### Operate without a desktop Check runs and logs, re-authorise connectors, and stop agents from a server or an SSH session. ### Let your coding agent build it npx -y @viorant/cli agents install teaches Claude Code, Codex or Cursor the whole flow: questions, a plan you approve, connect, build, deploy, one test run, then the logs. Deploys still run under that agent's own permission model. ## BUILT TO BE CHECKED ### Signed releases Every release is signed from the Viorant release workflow with a Cloud KMS key, and with Sigstore, recorded in the public Rekor transparency log. ### Verified on install On macOS and Linux, the installer checks the checksum and the release signature itself, no extra tools needed, and installs nothing if either fails. Install cosign for an independent check. On Windows, verify the sha256 checksum by hand. ### No provider keys stored Provider keys are sealed and sent to your deployment. A connector you sign in to before deploying waits in one owner-only file in the Space's git-ignored .vio folder until the deploy hands it to the agent. ### Run keys shown once A run key is printed once and kept in one owner-only file in your home folder, so vio run can call the agent. It never goes into the Space or your repository. Deploys run through [Helix](/helix), the same engine the Hub uses. ## COMING NEXT - vio as a coding agent: the same terminal, with a model behind it, metered on your credits. ## QUESTIONS ### Do I need the Hub? No. The Hub and vio deploy through the same pipeline. Use either. ### Does validating cost anything? No. Validation runs offline with no account. Deployed agents count against your plan's agent limit. ### Do I need my own provider key? No. vio deploy runs on your Viorant credits by default. Pass --env-file only if you want to bring your own key. ### Which coding agents does it work with? Claude Code, Codex and Cursor. npx -y @viorant/cli agents install writes the vio skill for them. Codex blocks network access by default: turn it on in its sandbox settings so vio can reach Viorant. ### How do I upgrade? Run the install command again, or vio update. npx -y @viorant/cli@latest always runs the current release. vio --version shows what you have. ## ONE DEFINITION. TWO WAYS TO SHIP IT. Author in the Hub or in your editor. Deploy from either. [Install vio](#install)[Download the Hub](/download)