| # | Tutorial | Description |
|---|---|---|
| 1 | Local HTTPS with a whoami service | Configure the gateway manually through the UI |
| 2 | Configure the Gateway with Owl AI | Let Owl AI do the configuration from a single prompt |
| 3 | Understanding the Dashboard | Read live traffic data and analyse request patterns with Owl |
| 4 | AI-Driven Protection — Disable & Re-enable a Service ← you are here | Use Owl to take a service offline under attack and bring it back |
Prerequisite: Complete Part 1 or Part 2 so you have
whoami.localrunning and accessible.
When a DDoS or targeted attack hits one of your services, the fastest response is to pull it offline at the gateway — no firewall rules, no server changes, just one command to Owl. The service goes dark in under a second. Once the attack subsides, Owl brings it back just as fast.
This tutorial demonstrates the full cycle: disable a cluster, verify the service is unreachable, then re-enable it.

Open the Owl chat panel (🦉 bottom-right) and type:
Disable the whoami cluster
Owl will call gateway_toggle_resource to disable the cluster. Aegis immediately rebuilds the xDS snapshot without it — Envoy stops routing to whoami within ~1 second.
Open https://whoami.local in your browser. You should see a 503 Service Unavailable — Envoy has no healthy upstream for the filter chain.
This is exactly what happens during a DDoS response: the service is shielded at the gateway layer while the backend stays untouched.
Once you're ready to bring it back:
Re-enable the whoami cluster
Owl toggles the cluster back on. Aegis pushes the updated snapshot to Envoy and https://whoami.local is accessible again within a second.
| Scenario | How Owl helps |
|---|---|
| DDoS targeting one service | Disable that cluster — attackers hit a 503, backend is protected |
| Vulnerability disclosed overnight | Owl can pull affected services offline while you patch |
| Maintenance window | Disable cleanly from chat, no SSH or YAML edits |
| Gradual re-enable after incident | Re-enable, watch dashboard, disable again if attack resumes |
The key advantage: this operates at the xDS layer. Disabling a cluster removes it from Envoy's routing table entirely — no lingering connections, no half-open sockets. The backend process keeps running; only the gateway path is cut.
Rather than waiting for you to notice an attack, Aegis can detect and respond automatically. When the AI patrol sweep identifies a high-threat pattern — flood of requests, known scanner signatures, spike from a single ASN — it can be configured to:
Go to Settings → AI to configure patrol sweep thresholds and notification channels.