SENTINEL COMMANDOperating layer
Every system, one operating view.
SEE. UNDERSTAND. AUTHORIZE. ACT. VERIFY.
Sentinel Command connects every system into one operational view. It is where a person sees the state of the estate and of Sentinel itself, understands it, authorizes a response, acts, and verifies the result.

- 01ObserveContinuous telemetry from every AXIS service and edge. (Command spans this stage)
- 02VerifyTelemetry checked at the gateway before it is trusted. (Command spans this stage)
- 03DetectAnomaly, policy violation, and absent signal alike. (Command spans this stage)
- 04ProtectPolicy enforced at the identity and tenant boundary. (Command spans this stage)
- 05PreserveEvidence written down before it can be lost. (Command spans this stage)
- 06RecoverContinuity coordinated, state restored, result verified. (Command spans this stage)
Recover returns to Observe
Command is where a person supervises every stage of the loop.
The operating picture
One view. Every system, each with its own state.
This is the Command demonstration's own surface, built from the same scenario. Step through it: detection, enforcement, preservation and a verified recovery.
The supervision cycle
See. Understand. Authorize. Act. Verify.
Command decides nothing on its own authority. It puts a person in front of every system's state, and sends a decision back out only under that person's recorded authority.



See
Watchtower reports a deployment as degraded and Guard reports a session it challenged. Both arrive in one view, each with its own state.
Understand
Trace places the two findings in sequence. Each record says how it is known: Watchtower's finding is server observed, the session report is site reported.
Authorize
Command offers only the responses the operator's role allows. The server re-validates the role, the tenant, the legal transition and the Safe Mode scope before anything happens.
Act
The operator contains the session and requests preservation. Guard enforces the containment, and the Ledger records both actions with the authority each used.
Verify
Watchtower confirms the deployment is reporting again and the recovery-verification task is complete. Only then does the server accept the incident's close.
Incident to recovery
From a declared incident to a verified restore point.
The same surface, the last four steps of the scenario: the incident, Safe Mode, COLOSSEUM preserving state, and the verified recovery. COLOSSEUM is a separate system; there is no live COLOSSEUM integration.
An incident is declared.
A commander declares the incident. The timeline now carries every correlated signal, the enforcement action and the sealed evidence reference — assembled as it happened rather than reconstructed afterwards.
- AXIS OMNIAserviceNo signal
- AXIS 7DserviceDegraded
- COLOSSEUMcontinuityReporting
- D1 INDEXevent indexReporting
- EVIDENCE STOREraw evidenceReporting
- AXIS COREserviceVerified
- CLOUDFLARE EDGEedgeVerified
- SENTINELcontrol planeVerified
Event stream
- Just nowhighAXIS-7DWorkspace Cedar
PRIVILEGED_SESSION_CREATED
- Just nowmediumSENTINELWorkspace Cedar
TRACE_ASSEMBLED
- Just nowmediumCLOUDFLARE-EDGEWorkspace Cedar
REQUEST_CORRELATED
- Shortly aftercriticalGUARDWorkspace Cedar
TENANT_BOUNDARY_VIOLATION
- Shortly afterhighGUARDWorkspace Cedar
SESSION_REVOKED
- ThenmediumEVIDENCE-STOREWorkspace Cedar
EVIDENCE_BUNDLE_SEALED
- ThenlowD1 INDEXWorkspace Cedar
EVENT_INDEX_COMMIT
- ThencriticalSENTINELWorkspace Cedar
INCIDENT_DECLARED
Simulated scenario on fabricated data. Every identifier, tenant and trace here is invented.
The intelligence theatre
Every system converges. Nothing is drawn that is not there.
The five Sentinel systems report into Command. The eight sectors beyond them are a design direction: this map gives each a place and no data, and says so when you select one.
Illustrative demonstration
- Built: reports to Command
- Design direction: no data connected
Core system · OBSERVE / DETECT
AXIS SENTINEL WATCHTOWER
Built
Visibility per deployment, the transitions and their reasons, and Watchtower health: the last scheduled success, the last status and whether it is stale.
Built: per-deployment heartbeat visibility evaluated every minute by a dedicated scheduled Worker, with its own run history and staleness.
States it reports
- Full coverage
- Degraded observation
- Blind spot
- Possible suppression
Sectors beyond the five core systems (security, infrastructure, financial, operations, vendors, compliance, external and AI systems) are a design direction for Command, shown only as labelled demonstrations.
Signal path
What enters it, and what leaves.
Signals in
- Each system's state through the authenticated API
- Events, incidents and their timelines
- Watchtower health, audit completeness and notification delivery
- The operator's role and tenant
Command
SUPERVISES THE ENTIRE LOOP
Outputs
- Triage transitions the server accepts
- Incident declarations and playbook progress
- Safe Mode engaged and lifted at a scope
- Authorised responses, recorded with their authority
Ten questions
Command, answered.
The questions every Sentinel system must answer, in the same order on every system page.
Sentinel Command is the operating view: one place where a person sees every system's state, understands it, authorizes a response, acts and verifies the result.
The operational state of the estate and of Sentinel itself.
- Events, incidents and their timelines
- Visibility and Watchtower health
- Evidence state, the Range gate and Safe Mode
- Continuity jobs, Shield records, team and authority
What each system reports through the authenticated API, scoped to the operator's own tenant and role.
None on its own authority. The console is static and holds no authority: every privileged decision is made server-side and re-validated there, so the interface can never grant what the API would refuse.
Triage events through a state machine whose legal transitions the server enforces, declare and work incidents from playbooks, engage and lift Safe Mode at a scope, request preservation, run Range scenarios and read the Ledger.
It cannot exceed the operator's role, reach another tenant, or show a sector it has no data for.
- The public Command demonstration runs on fabricated data and is labelled as a demonstration.
Design directionSectors beyond the five core systems (security, infrastructure, financial, operations, vendors, compliance, external and AI systems) are a design direction for Command, shown only as labelled demonstrations.
Command is the destination the other systems report to. It reaches outward through notifications, and their delivery is itself tracked and shown.
Every core system reports into Command, and Command sends decisions back out only under the operator's recorded authority: a Safe Mode engagement, a triage transition, a preservation request.
A stale source is shown as stale. Watchtower's staleness, pending audit writes and undelivered notifications are surfaced beside the estate's state, never folded into a calm picture.
Command's reads are the server's answers and are never cached. Watchtower health, audit completeness and notification delivery are each shown as separate facts.
Relations
Where it connects.
- Guard Receives fromDecisions and containment state
- Watchtower Receives fromVisibility and Watchtower's own health
- Range Receives fromNamed results and the release gate
- Evidence Receives fromPulse, Trace and the integrity of the record
- Continuity Receives fromPreservation jobs and recovery verification
- Safe Mode Sends toEngage and lift, by operator authority
- The operator Exchanges with, outside SentinelState shown; decisions taken within the operator's role
- Sends to
- Receives from
- Exchanges with
- Outside Sentinel
Failure and health
When it fails, it says so.
In operation
Current
- Reads are the server's answers
- Never cached
When it fails or loses visibility
Stale source
- Watchtower staleness shown beside the estate
- Pending audit writes surfaced
- Undelivered notifications surfaced
Command states
- NominalFull observation. Nothing open.
- ObservingOpen findings under routine work.
- DegradedTelemetry incomplete. Visibility is not assured.
- InvestigatingA declared incident is being worked.
- ContainedSpread halted. Not yet resolved.
- Safe ModeDegraded deliberately. High-risk operations held.
- CriticalActive critical condition requiring response.
- Current (lit: in operation)
- Stale source (lit: on failure)
How its own health is verified
- Each health fact shown separately
- Authority re-validated server-side
- Triage transitions enforced by the server
The Sentinel system family
