All it needs: read-only Kube creds and outbound HTTPS to api.byteport.polsia.app. In the first 5 minutes it inventories signals, classifies incidents by type, and dry-runs every fix before executing.
Default mode is dry_run — it plans every action and posts it to your ops channel. Nothing executes without approval for the first 24 h.
api.byteport.polsia.app:443 — used only to report runbook decisions and post-remediation postmortems. No inbound port needed.| Resource | Verbs | Why |
|---|---|---|
| pods, nodes, events | get, list, watch | Signal collection |
| deployments, replicasets | get, list, watch | Runbook matching |
| deployments | patch | Rollback action only — guarded by allowlist |
| pods/exec | create | Diagnostics (df -h, free -m) — read-only commands |
Two options — pick whichever fits your workflow. Both produce identical runtime behaviour.
Option A — Helm (recommended)
# Add the BytePort chart repo
helm repo add byteport https://byteport.polsia.app/charts
helm repo update
# Install — default mode is dry_run (safe to run in prod immediately)
helm install byteport byteport/byteport-agent \
--namespace byteport-agent \
--create-namespace \
--set config.apiKey=YOUR_BYTEPORT_API_KEY \
--set config.clusterName=prod-us-east-1 \
--set config.mode=dry_run
# Confirm the agent is running and watching (logs appear within 30s)
kubectl get pods -n byteport-agent
kubectl logs -n byteport-agent -l app.kubernetes.io/name=byteport-agent --tail=20Option B — Raw manifest
# Download and inspect the manifest before applying
curl -sL https://byteport.polsia.app/install/manifest.yaml -o byteport-manifest.yaml
# Apply to your cluster
kubectl apply -f byteport-manifest.yaml
# Set required env vars on the Deployment
kubectl set env deployment/byteport-agent \
-n byteport-agent \
BYTEPORT_API_KEY=YOUR_API_KEY \
BYTEPORT_CLUSTER_NAME=prod-us-east-1 \
BYTEPORT_MODE=dry_runAfter install, you should see the agent reporting to your ops channel within 60 seconds:
byteport-agent INFO Signal monitor started — cluster: prod-us-east-1
byteport-agent INFO Adapter: kubernetes-events connected (12 nodes, 47 pods)
byteport-agent INFO Mode: dry_run — runbook plans will be posted to Slack before execution
byteport-agent INFO Ready. Watching for: disk_pressure, memory_pressure, pod_crashloop, tls_cert_expiring, db_connections_saturatedThree required vars. Everything else is optional.
# Required
BYTEPORT_API_KEY=bp_live_... # from /pricing — Starter is free
BYTEPORT_CLUSTER_NAME=prod-us-east-1 # shows up in postmortems + Slack messages
BYTEPORT_MODE=dry_run # dry_run | auto (see below)
# Notify — at least one required (Slack recommended)
BYTEPORT_SLACK_WEBHOOK_URL=https://hooks.slack.com/services/...
BYTEPORT_PAGERDUTY_ROUTING_KEY=... # optional — escalation fan-out
# Signal adapter — defaults to Kubernetes events; override for Prometheus/Datadog
BYTEPORT_ADAPTER=kubernetes # kubernetes | prometheus | datadog | cloudwatch | sentry
PROMETHEUS_URL=http://prometheus:9090 # if adapter=prometheusauto is a one-liner: helm upgrade byteport byteport/byteport-agent --set config.mode=auto.
Here's the full loop from alert to postmortem — takes about 4 minutes for a disk-pressure incident:
To trigger a synthetic test signal without waiting for a real incident:
# Trigger a synthetic disk_pressure signal (safe — dry_run by default)
curl -X POST https://byteport.polsia.app/api/webhooks/signal \
-H "Content-Type: application/json" \
-d '{
"signal_type": "disk_pressure",
"value": 91.5,
"severity": "warning",
"cluster": "prod-us-east-1",
"node": "node-3"
}'Check byteport.polsia.app/incidents within 30 seconds — your signal will appear in the live feed.
After install, BytePort auto-handles 9 pager classes without intervention. See the full autonomy matrix →
The #1 question from security-conscious buyers. Here's the complete list:
One command. Removes all cluster resources including the ServiceAccount and RBAC roles:
helm uninstall byteport --namespace byteport-agent
kubectl delete namespace byteport-agentSignal history and postmortems are kept in your BytePort account for 90 days after uninstall. Data deletion on request.