Open-source AI Security Operations Center: alert fusion, LLM-agent triage, MITRE ATT&CK investigation, and a replayable decision ledger for every agent step. Self-hostable, runs with no API keys, MIT licensed. Ships an MCP server for Claude, Cursor and Continue.
AiSOC incident runbooks
This directory holds the on-call playbook for every alert defined in
infra/docker/alerts/aisoc.rules.yml.
Each runbook follows the same five-section structure so an on-call
engineer can locate the right diagnostic command without reading
prose end-to-end:
- What the alert means — one paragraph; what symptom is Prometheus describing.
- First five minutes — copy-pasteable commands that turn the alert into a hypothesis. Logs, metrics queries, container status. No prose.
- Mitigation — what to do BEFORE root-causing. The point is to stop customer pain even if we don't yet know why it started.
- Root cause — the deeper debugging path once the bleeding has stopped.
- References — source code, dashboards, Slack channels.
Runbook quality bar: a runbook is "done" when an engineer who has never seen this alert before can resolve a real incident using only the runbook and the standard AiSOC dev environment. If you ever followed the runbook and it didn't work — fix it before you close the incident, not after.
Index
| Alert | Severity | Runbook |
|---|---|---|
AisocServiceDown |
critical | service-down.md |
AisocHttpP95LatencyHigh |
warning | http-latency-high.md |
AisocHttpErrorRateHigh |
critical | http-5xx-high.md |
AisocKafkaConsumerLagHigh |
warning | kafka-consumer-lag.md |
AisocDetectionPipelineStalled |
critical | detection-pipeline-stalled.md |
AisocActionExecutorFailureRateHigh |
critical | action-executor-failures.md |
| (not alert-bound — manual triage) | — | database-incident.md |
Versioning
Updated: 2026-06-28 (Phase 2.5).