Monitoring AI coding agent costs and activity on your own machine

Description and use

Developers using multiple AI agents now have a unified dashboard to monitor costs, token usage, and session activity, enabling faster debugging and cost optimization.

Problem

Developers using multiple AI coding agents lacked a unified view of costs, token usage, and session activity, making it hard to diagnose slow or failed sessions.

Outcome

Developers gain real‑time visibility into AI agent costs and activity, enabling faster diagnosis of failures and optimization of spending.

Workflow

• Install AgentMon Start on Windows/macOS/Linux
• Sign in via Codenotary
• Collector parses agent logs into snapshots
• Local server stores snapshots in SQLite
• Dashboard visualizes costs, token usage, tool calls, errors, and security findings
• Use session view to inspect threads, tool calls, and failures
• Use trace waterfall to debug stalled or unexpected sessions
• Set alerts for cost, error, and activity thresholds
• Review recommendations for cost and reliability improvements

Evidence

I run Claude Code, Cursor, Codex and a handful of sub-agents on my own machines every day.

Source

AgentMon Start reads the session logs your coding agents already write.

Source

The Overview shows spending alongside session counts, token usage and cache hit rate.

Source

The Tools view also aggregates failures across sessions, so you can see whether one tool accounts for most of the errors.

Source

Alerts check every 60 seconds. Depending on your plan, they can flag cost thresholds, spending spikes, error spikes, long-running sessions and unmanaged agent activity.

Source