Deployments overview
A deployment takes something you built in the Studio — an agent, a multi-agent system, a copilot, or a vibe-coded app — and makes it live for real users. On the managed cloud that means a public URL under qraptor.app.
The Deployments section of a project is where you manage those live, production deployments: their agents, schedules, API keys, custom domains, monitoring, and settings.
The managed cloud is not the only place a project can run. If it has to run somewhere you control, see Where a project can run at the end of this page.
Preview vs production
There are two ways your work goes live, and they are not the same thing.
| Preview | Production | |
|---|---|---|
| What it is | A temporary environment for testing while you build | A stable, always-on deployment for real users |
| Where it lives | Inside the builder (the app, copilot, or agent editor) | The Deployments section |
| Lifetime | Ephemeral — stops automatically when idle | Always-on until you delete it |
| URL | A temporary preview link | your-app.qraptor.app or a custom domain |
| Use it for | Internal testing and sharing with your team | End-user traffic |
Preview is a build-time step, so it is documented with each builder:
Everything else on this page — and in this section — is about production.
Always validate behavior in preview before you deploy to production. A production deployment serves real users and counts against your plan’s rate limits and credits.
What can be deployed
A production deployment is one of four types, depending on what you are shipping:
- Web App — a full vibe-coded application (frontend + backend).
- Copilot — an embeddable AI chat widget.
- Agent API — one or more agents exposed as API endpoints, with no user-facing page.
- Multi-Agent — a whole multi-agent system behind a single endpoint.
Agents can also be layered on top of a Web App or Copilot deployment. See Deployment types for what each one provisions and when to choose it.
Where deployments are created
You start a deployment from wherever you built the thing you are shipping:
| To deploy a… | Start from | Guide |
|---|---|---|
| Web App | The vibe workspace | Publish to production |
| Copilot | Copilot Studio | Deploy a copilot |
| AI Agent | The agent editor’s Deploy menu | Deploy an AI Agent |
| Task Agent | The agent editor’s Deploy menu | Deploy a Task Agent |
| Agent API / Multi-Agent | The Deploy menu on the Deployments page | Production deployments |
The Deployments page
Open Deployments from the project sidebar to see every production deployment in the project. Each one appears as a card showing its type, status, primary URL, what’s deployed inside it, and how it’s authenticated.
Deployment status
| Status | Meaning |
|---|---|
| Provisioning | The deployment is being set up. |
| Ready | Live and serving traffic. |
| Degraded | Live but reporting errors or health issues — check monitoring. |
| Stopping | Being torn down. |
| Stopped | Not serving traffic. |
| Failed | Provisioning did not complete. |
What you can manage
Open a deployment to reach everything below. The available sections depend on the deployment type.
- Deployment types — Web App, Copilot, Agent API, and Multi-Agent.
- Production deployments — create, update, and understand a running deployment.
- Managing agents — versions, rate limits, testing, and agent-to-agent access.
- Multi-agent systems — deploy a whole system behind one endpoint.
- Scheduled runs — run an agent on a recurring schedule.
- Database triggers — run an agent when a row changes.
- API access — API keys and calling your deployment programmatically.
- Monitoring — metrics, traces, executions, and conversations.
- Custom domains — serve a deployment from your own domain.
- Settings — authentication and deleting a deployment.
A deployed agent still runs on qRaptor’s infrastructure. If some of its steps need to reach a private database or an internal API, those individual steps can be routed to a remote executor running inside your own network — the same routing applies to deployed agents as to ones you’re testing in the builder.
Where a project can run
The managed cloud is the default, not the only option. There are three shapes, and they differ in how much runs on your side:
| Managed cloud | Remote executor | Exported appliance | |
|---|---|---|---|
| Runs the platform | qRaptor | qRaptor | You |
| Runs the work | qRaptor | Individual steps run in your network | You |
| Where data goes | qRaptor cloud, with residency options on Enterprise | Sensitive calls stay inside your network | Never leaves your infrastructure |
| You manage infrastructure | Nothing | A lightweight executor | The whole appliance |
| Use it when | The default | Agents must reach a private database or internal API | The system has to run somewhere you control |
- Remote executors — keep the platform managed and push only the sensitive steps into your network.
- Export as an appliance — take the whole governed system off-platform as a sealed, signed archive.