> ## Documentation Index
> Fetch the complete documentation index at: https://docs.viorant.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Deploy an agent

> Take an agent off your machine and give it a URL.

Running an agent [locally](/run-agent-locally) needs the Hub open. **Deploying** it puts
the agent on Viorant Cloud, where it keeps running whether the Hub is open or not, and
answers on its own URL.

## Before you deploy

Two things have to be true.

**The agent needs a `.vio` bundle.** That's the portable file describing the whole agent —
prompt, model, skills, connectors and memory — as one thing. Use **Create `.vio`** on the
agent, or **Update `.vio`** if you've changed the agent since the last one.

**The agent must run on your own provider key.** Deployed agents can't spend Viorant
credits, so an agent set to the Viorant provider is refused before anything is created.
Point it at a provider you've added a key for — see [Bring your own key](/providers/byok).

<Note>
  `.vio` is built from your **committed** agent, not your working edits. If **Update
  `.vio`** is unavailable, commit the agent first.
</Note>

## Deploying

<Steps>
  <Step title="Open the agent and choose Deploy">
    **Deploy** sits beside **Create `.vio`** on the agent editor.
  </Step>

  <Step title="Name it and pick where it runs">
    The name is how you'll recognise it later. Deploying the same agent twice under
    different names is a normal thing to do — that's how you keep a stable one and a
    test one side by side.
  </Step>

  <Step title="Confirm the credentials it needs">
    The wizard lists the provider keys and connector accounts the agent depends on.
    These are encrypted on your machine and uploaded directly to the deployment —
    Viorant never sees them in the clear.
  </Step>

  <Step title="Copy the run key">
    Shown **once**, at the end. It cannot be retrieved later from the create screen.
    Store it somewhere safe before closing.
  </Step>
</Steps>

## Calling it

Each deployment gets a URL and a run key. The key goes in the `Authorization` header:

```bash theme={"dark"}
curl -X POST https://<your-deployment-url>/run \
  -H "Authorization: Bearer <run-key>" \
  -H "Content-Type: application/json" \
  -d '{"inputs": {"topic": "quarterly summary"}}'
```

The wizard shows a ready-made `curl` for your specific deployment, with its own input
names filled in.

## What you can deploy

Each plan allows a number of running agents at once. If you're at the limit, stop one
before deploying another — see [Manage deployments](/deploy/manage-deployments).

<Warning>
  An abandoned deployment still counts against your limit until it's stopped. If a deploy
  fails partway and you close the wizard, cancel it rather than leaving it.
</Warning>

<Card title="Manage deployments" icon="sliders" href="/deploy/manage-deployments">
  See runs, check status, and stop an agent.
</Card>