Operations Center
The Operations Center (/ops) surfaces items requiring operator action, prioritized by severity. Failed runs, budget breaches, gateway outages, and escalated approvals all appear here.
| Action | How |
|---|---|
| Triage | Items sorted by severity (critical first, then warning, then info) |
| Acknowledge | Click to suppress an item temporarily. Reappears if the condition persists |
| Add notes | Attach context for other operators investigating the same issue |
| Navigate | Each item links directly to the source resource (run, agent, gateway) |
Attention Item Types
Critical
Gateway offline, infrastructure failures — affect all agents on the gateway
Warning
Run failures, budget pauses, gateway degradation, policy violations
Info
Aging approvals, unclassified failures
Correlated Changes
6-hour lookback for deployments, config changes, and policy updates that may explain the issue
Event Timeline
The event timeline (/observe) provides a unified view of all platform events. It merges Convex platform events with Bridge SSE into a single filterable stream.
- Filter by category — executions, governance decisions, lifecycle transitions, security events
- Filter by time range — 5m, 15m, 1h, 24h, or all
- Filter by resource — scope to a specific run, agent, or gateway
- Filter by actor — see all actions by a specific user or agent
- Search — free-text search across event payloads
Traces
Step-level execution traces for debugging workflow runs. For each step, the trace shows:| Detail | Description |
|---|---|
| Timing breakdown | Queued, started, completed timestamps with duration |
| Conversation log | The agent’s full message transcript for that step |
| Tool calls | Every tool invocation with arguments and results |
| Outputs | The step’s produced outputs and artifacts |
| Failure detail | Error message and stack trace for failed steps |
Copilot Analysis
AI-powered run analysis for failed or anomalous runs. When a run fails or exhibits unusual behavior, trigger Copilot analysis from the run detail page. Copilot examines the execution trace — step sequence, tool calls, errors, timing — and suggests root causes. Analysis runs in a dedicated thread and produces a structured report with findings and suggested next steps. Useful when the failure isn’t obvious from the trace alone.Operator Notes
Attach notes to any resource — a run, an agent, a factory. Notes are shared across the team for institutional context and persist across sessions. Use notes for:- Post-mortem annotations on failed runs
- Configuration rationale on agent settings
- Handoff context when transferring operational responsibility