Monitoring

The monitor is Stagent’s real-time execution feed. It complements the inbox by showing what agents are doing right now, not just the points where they need human input.

Monitor with 3 active agents, 92% success rate, tasks today count, and streaming log panel

Key Concepts

  • Workspace Health Cards — Active agents, tasks today, success rate, and last activity stay visible above the log stream.
  • Streaming Log View — Runtime events appear chronologically with sub-second latency so operators can follow one task or monitor the whole system.
  • Task & Event Filters — Narrow the stream to one task set or one class of event (reasoning, tool calls, tool results, decisions, errors) when debugging or demoing the product.
  • Click-to-Navigate — Click any log entry to jump directly to the task detail view for deeper inspection.
  • Auto-Pause on Tab Switch — Streaming pauses when you switch browser tabs to save resources, then resumes and backfills missed events when you return.
  • Connection Status — The monitor makes it explicit whether the stream is connected and whether auto-refresh is enabled.
  • Auto-Pause on Tab Switch — Streaming pauses when you switch browser tabs to save resources, then resumes and backfills missed events when you return.

Log Entry Types

Each log entry includes:

  • Timestamp — When the event happened
  • Task Context — Which task or workflow step produced the event
  • Event Type — Tool call, content block, completion, or other runtime marker
  • Content — The event payload or summarized runtime output

How It Works

The monitor is the fastest way to understand whether work is progressing normally. Operators can keep it open during a run, then pivot to the inbox only when the agent crosses a trust boundary or needs clarification.

Because the monitor is filterable, it also works as a debugging surface. You can focus on one task, isolate one event class, and inspect the sequence that led to a failure or approval request.

Use Cases

Live Demo for Stakeholders

Product Team · Showcasing agent capabilities to executives

You project the monitor during a product demo. Stakeholders see an active log stream, success-rate card, and visible task progress instead of a black-box spinner.

Debugging a Failed Task

Developer · Tracing a failure to its root cause

A nightly task failed. You filter the log stream to the relevant task, inspect the final tool calls, and confirm whether the problem was a permissions denial, missing file, or execution error before retrying.

Overnight Batch Monitoring

Power User · Morning triage of unattended agent work

You open the monitor after a night of scheduled work and can immediately tell whether the system is healthy, which task is still active, and where intervention is needed.