Reality probe against cloudflared
INFRA.md says "22 subdomains". Reality probe runs cloudflared list-tunnels weekly, finds 25 hostnames. Drift proposal appears with the exact diff of the 3 new hostnames.
All verticals
schema_pack · ops
Ops and infrastructure
Runbooks, postmortems and infra docs that stay alive, validated against reality.
Ops teams maintain runbooks, postmortems and infra docs. The classic problem: docs age, diverge from reality, and whoever arrives on-call finds out the hard way. ValorBrain combines canonical sources with reality probes: the canonical doc is periodically re-validated against the actual system.
What changes
Same job, with and without organizational memory.
Without
With ValorBrain
A January runbook describes a flow that changed in March. found out during an incident
Schema pack "ops" structures: runbook, postmortem, sla, dependency, oncall, infra_change
Postmortems get lost in the docs folder and are not consulted on similar incidents
Reality probes validate canonical docs against reality (cloudflared, systemd, k8s)
Architecture diagram out of date for months and nobody notices
Drift detection flags when INFRA.md diverges from real state
On-call rotation changes but the "responsibilities" doc stays old
Postmortems indexed with causes + actions, surface automatically on similar incidents
—
Cross-agent: Hermes ingests logs, Claude Code writes the fix, Kiro records the postmortem
In practice
Concrete paths teams already run.
Postmortem from another team
On-call opens an incident similar to one already resolved by another engineer 4 months ago. ValorBrain surfaces the old postmortem automatically with the applied fix and resolution time.
Validated runbook
Runbook says "run systemctl restart valorbrain-watcher". Reality probe confirms the service exists and is enabled. If the name changes, drift detects it before it becomes an incident.
Schema pack ops
Observation types the decision-extractor captures in this domain.
| type | captures |
|---|---|
runbook | Procedures with prerequisites, steps, validation |
postmortem | Incidents with root cause, timeline, action items |
sla | SLAs with targets and real measurements |
dependency | Service dependencies with health checks |
oncall | Rotations with current owners and contacts |
infra_change | Changes applied with date, author, impact |
Agents
MCP-compatible clients at mcpbrain.valor.digital.
- Claude Code
- Hermes
- Kiro CLI
- agent custom monitoring
Run this for ops and infrastructure
Minutes to connect a source. Free credit on signup.