By name: oncall description: Control the OnCall Agent demo - activate monitoring, check investigation status, approve/deny credentials, simulate anomalies, and manage incident response. Use when the user mentions oncall, incidents, monitoring, production issues, anomalies, or wants to interact with the demo.
OnCall Agent Demo
This skill helps you interact with the OnCall Agent - an AI-powered incident response system that monitors infrastructure, detects anomalies, and orchestrates remediation with secure credential access via 1Password.
When to Use
- User wants to activate or start the OnCall agent
- User asks about production status, incidents, or what's happening
- User wants to check investigation progress or root cause
- User needs to approve or deny credential requests
- User wants to simulate an anomaly for demo purposes
- User wants to reset the demo
Environment
The OnCall demo is deployed at:
- Dashboard URL: https://a9eb812238f753132652ae09963a05e9-oncall.cluster.autonomy.computer/
- API Base: https://a9eb812238f753132652ae09963a05e9-oncall.cluster.autonomy.computer
Set ONCALL_URL environment variable to override.
Demo Flow
- Inactive → Activate the agent
- Activation Approval → Approve READ credentials (5 credentials from 1Password)
- Monitoring → Agent watches infrastructure; simulate anomaly for demo
- Incident Detected → 150 diagnostic agents deploy across 3 regions
- Investigating → Agents analyze services in parallel
- Diagnosis Complete → Root cause identified
- Write Approval #1 → "Kill 3 runaway queries" - typically DENIED
- Write Approval #2 → "Disable feature flag" - typically APPROVED
- Remediating → 30-45 second remediation with progress
- Resolved → Incident closed, audit trail available
- Reset → Return to Inactive state
Scripts
Use these scripts to interact with the OnCall API:
Check Status
Run scripts/status.sh to get the current agent status and any pending approvals.
Activate Agent
Run scripts/activate.sh to start the OnCall agent (requests READ credentials).
Approve Credentials
Run scripts/approve.sh to approve pending credential requests.
Deny Credentials
Run scripts/deny.sh to deny pending credential requests.
Simulate Anomaly
Run scripts/simulate.sh to trigger a cascading failure anomaly (demo only).
Reset Demo
Run scripts/reset.sh to reset the demo to the inactive state.
API Reference
GET /oncall/status
Returns current agent state including:
status: Current state (inactive, monitoring, diagnosing, etc.)waiting_for_approval: Boolean indicating if approval is neededapproval: Details about pending credential requestsagent: Agent-specific data (diagnosis results, remediation progress)
POST /oncall/activate
Starts the OnCall agent. Returns status waiting_for_activation_approval.
POST /oncall/approve
Body: {"approved": true} or {"approved": false}
Approves or denies pending credential requests.
POST /oncall/simulate-anomaly
Triggers a cascading failure anomaly for demo purposes. Only works when status is monitoring.
POST /oncall/reset
Resets the demo to inactive state.
GET /graph
Returns the agent visualization graph with nodes and edges.
Interpreting Status
| Status | Meaning |
|---|---|
inactive |
Agent not running, ready to activate |
waiting_for_activation_approval |
Needs READ credential approval |
activating |
Retrieving credentials |
monitoring |
Actively watching infrastructure |
incident_detected |
Anomaly found, deploying agents |
diagnosing |
Investigation in progress |
diagnosis_complete |
Root cause identified |
waiting_for_write_approval |
Needs WRITE credential approval for remediation |
remediating |
Executing approved fix |
resolved |
Incident closed successfully |
Tips
- Always check status before taking actions
- The dashboard provides real-time visualization of agents
- Watch the graph fill with gold dots (deploying) turning green (complete)
- The first WRITE approval (kill queries) is designed to be denied
- The second WRITE approval (disable feature flag) is designed to be approved
- Remediation takes 30-45 seconds with visible progress