Agent Session Monitor
A read-only monitoring tool that brings coding-agent sessions into one view, with activity, attention signals, and evidence about what is actually happening.
Screenshot unavailable. The project description below explains the workflow.
Know which agent needs you.
When several coding sessions are running, the hard part is knowing where attention is needed. This monitor discovers local sessions and exposes their observed state through a CLI, dashboard, and read-only MCP tools.
How it works
Each state includes its evidence source, freshness, and confidence. The monitor stores metadata rather than conversation content. An idle process is not treated as proof that the task is finished.
Workflow example
- Discover local sessions across the supported coding providers.
- Filter for sessions waiting on approval, input, or other attention.
- Inspect the evidence and return to the original session to act.
Project scope
This is a local monitoring tool. The portfolio shows its operating model without exposing live sessions or conversation data.