Deploy

By the end of this page your system will be live at its own URL, callable with an API key.

Prerequisites

• A production version of the system (Publish).
• Every member agent published to production (Publish an AI Agent).
• Admin role.

Two ways in

Both open the same dialog: Deploy Multi-Agent System.

Deploy it

Check the system summary

The top of the dialog confirms what you’re shipping: the system name, a Production badge, the Conversation Router pattern, and the member count. If it reads No Production Version, publish first — the dialog tells you so and won’t let you continue.

Choose a target

• New deployment — creates a new endpoint with its own URL and API keys.
• Existing deployment — adds the system to a deployment you already have, sharing its URL and API keys. Only multi-agent deployments are eligible.

Name it

For a new deployment, set the App Slug. Availability is checked as you type, and the dialog previews the resulting endpoint:

your-slug.qraptor.app/mas/{systemId}/conversation/stream

Production domain offers qraptor.app (default) or the trusted domain app.qraptor.ai — useful when a customer’s network blocks qraptor.app. The choice is saved to the project.

Leave API Keys Enabled on unless you have a reason not to; it’s how you’ll authenticate calls.

Set the rate limit

Rate Limit (requests/min) caps traffic to the system’s endpoint. The default is 60.

Review the members

Member Agents lists every specialist with an Auto-deployed badge. You don’t add them separately — deploying the system deploys them, and removing the system removes them again.

Deploy

Click Deploy System. Review the credit summary shown in the dialog before you confirm.

On success you get the API Base URL and, for a new deployment with API keys enabled, a one-time API key. Copy the key now — it can’t be retrieved later. If you miss it, generate another from the deployment’s API Keys tab.

Calling the deployed system

Send conversation turns to the system’s endpoint, authenticated with an API key:

curl -N https://your-slug.qraptor.app/mas/{systemId}/conversation/stream \
  -H "Authorization: Bearer qr_your_api_key" \
  -H "Content-Type: application/json" \
  -d '{"message": "What was total revenue on 2026-07-20?"}'

The router classifies each turn and delegates to a member exactly as it does in the builder preview. Replies stream back as server-sent events.

See API access for key scopes, rotation and revocation.

After deploying

The deployment gains a Systems tab for the system itself — its endpoint, its pinned version, and its members — and its member agents appear on the Agents tab with their own test console and endpoints.

To roll out a new version of the system, publish it here, then switch the version on the deployment’s Systems tab. See Multi-agent systems for the full operational picture.

Common issues

• Deploy is greyed out in the builder — No production version yet. Publish first.
• “Cannot deploy system: member agent … has no published version.” — Publish the named agents, then retry.
• “…is already deployed on this deployment” — A member is already deployed there on its own. Remove it from that deployment first, or deploy the system elsewhere.
• The slug is taken — Slugs are globally unique. Pick another; only lowercase letters, numbers and hyphens are allowed.
• No existing deployments listed — There’s no multi-agent deployment yet. Create a new one instead.
• The deployed system behaves differently from the preview — The deployment runs its pinned version, not your current draft. Publish, then update the version on the Systems tab.

Next

• Multi-agent systems — manage, version and monitor the deployment.
• API access — API keys and calling conventions.