Production deployments
By the end of this page you will have created a production deployment, know how to read a running deployment’s dashboard, and understand how updates roll out without downtime.
Prerequisites
- Starter or higher.
- Admin role.
- For an Agent API or Multi-Agent deployment: at least one agent (or system) with a PRODUCTION version. See Publish an AI Agent or Publish a Task Agent.
Creating a deployment
How you start depends on the deployment type:
- Web App and Copilot deployments are created from their builders — see Publish to production and Deploy a copilot.
- Agent API and Multi-Agent deployments are created from the Deploy menu on the Deployments page.
The rest of this page walks through creating an Agent API deployment. Multi-agent systems follow the same pattern.
Open the Deploy menu
On the Deployments page, click Deploy and choose Agent. The deploy wizard opens on the Target step.
Name the deployment
Choose Create new agent-only deployment and enter a Deployment Slug — this becomes the deployment’s URL. As you type, the Studio checks that the slug is available and previews the full URL.
You can also pick the production domain the slug sits under. The default is qraptor.app. To add your own domain instead, deploy first, then follow Custom domains.
To add agents to a deployment that already exists, choose Add to existing deployment on this step instead of creating a new one. You can add agents to any deployment type.
Select the agents
On the Agents step, pick the agents to include. Only agents with a PRODUCTION version appear.
Configure each agent
On the Configure step, set each agent’s version behavior (latest production or pinned) and rate limit. These options are explained in Managing agents.
Review and deploy
The Review step shows the agents and a credit summary, plus an optional Generate API key toggle. Deploy, and the deployment provisions and moves to Ready.
The wizard’s key toggle just generates a key (shown once on the success screen) with a default scope set — invoke + runs:read, plus a2a if any included agent has A2A enabled. To create a key with a specific name, a single scope, and an expiry, use the deployment’s API Keys tab after it’s ready.
The deployment dashboard
Open a deployment to land on its Dashboard — a summary of health, recent usage, what’s deployed, and how to reach it.
The dashboard shows:
- Health — status, current version, and 24-hour request and error rate.
- Quick metrics — requests, error rate, P95 latency, and unique users over the last 24 hours.
- What’s deployed — the application service (for user-facing types) and the deployed agents.
- Access — the URL or API base URL, and the authentication methods in use.
Updating a deployment
You update a production deployment in place — there’s no separate redeploy for most changes:
- Adding or removing agents takes effect immediately. See Managing agents.
- New agent versions roll out automatically. When an agent is set to use its latest production version, promoting a new version switches the deployment to it — you don’t redeploy. (If you pinned the agent to a specific version, it stays on that version until you change the pin.)
- Authentication changes — toggle OIDC or API keys in Settings.
Updates are applied as a rolling, zero-downtime change — traffic keeps flowing to the current version until the new one is ready.
Runtime and tiers
- Agent API and Multi-Agent deployments run serverless — there’s no dedicated pod to manage.
- Web App and Copilot deployments run on dedicated capacity sized to your plan.
Rate limits scale with your plan.
Common issues
- Deploy is disabled or the agent isn’t listed — the agent needs a PRODUCTION version first. See the agent’s Publish page.
- Slug is unavailable — slugs are unique across the platform. Pick another, or use a custom domain.
- Stuck in Provisioning — give it up to a minute. If it doesn’t reach Ready, check the activity feed for the failing step.
- API calls return 401 — use a valid API key or token for the deployment. See API access.