Claude Code Monitoring provides end-to-end visibility into Claude Code CLI usage using its built-in OpenTelemetry telemetry output. The included dashboard helps engineering and AI leaders understand:
- Spend and token consumption over time
- Session volume and active time
- API request health (errors, latency, volume)
- Tool activity (tool calls, permission decisions, success rate, duration)
- Engineering output indicators (commits, PRs, lines added/removed)
What you’ll see (dashboard tiles & why it matters)
The dashboard summarizes
Total Users, Total Cost (USD), Total Tokens, Total Sessions, and Active Time to quantify adoption and spend.
It shows cost over time and token usage over time to spot spikes and trends.
It breaks down event volume by type (api_request, api_error, tool_decision, tool_result, user_prompt) to understand where activity and failures occur.
Engineering output signals include
- Lines of Code Added/Removed, Total Commits, Total PRs, plus LoC over time and PRs over time to correlate usage with delivery.
API analytics cover
- Cost by model, token & cost by model, and API latency over time (avg duration) to pinpoint expensive or slow model usage.
Tooling insights include:
- Most used tools (e.g., bash/read/edit/write), avg tool duration, permission decisions (allow/deny), and tool success rate (success/failure) to validate safety and reliability.
User activity include:
- Avg session time, avg cost per session, avg tokens per request, and sessions by user for adoption and accountability.
A recent event stream
- Supports debugging and drilldown of prompts, tool calls and session information
Claude Code Agent monitoring
