diff --git a/AGENTS.md b/AGENTS.md
index 4dc515a..2ea64ba 100644
--- a/AGENTS.md
+++ b/AGENTS.md
@@ -928,25 +928,24 @@ Continue on gb/ workspace. Follow AGENTS.md strictly:
      "sudo incus exec system -- stat -c '%y' /opt/gbo/bin/botserver"
    ```
 
-3. **Restart in prod** — after binary updates:
+3. **Restart in prod** — CI/CD handles this automatically:
    ```bash
-   ssh -o StrictHostKeyChecking=no -o ConnectTimeout=10 <PROD_HOST> \
-     "sudo incus exec system -- pkill -f botserver || true"
-   sleep 2
-   ssh -o StrictHostKeyChecking=no -o ConnectTimeout=10 <PROD_HOST> \
-     "sudo incus exec system -- bash -c 'cd /opt/gbo/bin && RUST_LOG=info nohup ./botserver --noconsole > /opt/gbo/logs/stdout.log 2>&1 &'"
+   # CI deploy workflow:
+   # Step 1: Check binary
+   # Step 2: Backup old binary (cp /opt/gbo/bin/botserver /tmp/botserver.bak)
+   # Step 3: Stop service (systemctl stop botserver)
+   # Step 4: Transfer new binary (tar+gzip via SSH)
+   # Step 5: Start service (systemctl start botserver)
    ```
 
 4. **Verify deployment**:
    ```bash
-   # Wait for bootstrap (~2 min)
-   sleep 120
-   # Check health
-   ssh -o StrictHostKeyChecking=no -o ConnectTimeout=10 <PROD_HOST> \
-     "sudo incus exec system -- curl -s -o /dev/null -w '%{http_code}' http://localhost:8080/health"
-   # Check logs
-   ssh -o StrictHostKeyChecking=no -o ConnectTimeout=10 <PROD_HOST> \
-     "sudo incus exec system -- tail -30 /opt/gbo/logs/stdout.log"
+   # Check service status
+   ssh <PROD_HOST> "sudo incus exec system -- systemctl status botserver"
+   # Monitor logs
+   ssh <PROD_HOST> "sudo incus exec system -- journalctl -u botserver -f"
+   # Or check stdout log
+   ssh <PROD_HOST> "sudo incus exec system -- tail -30 /opt/gbo/logs/stdout.log"
    ```
 
 ### Production Container Architecture