Cc Metrics

Claude Code Metrics Dashboard

A comprehensive monitoring stack for tracking Claude Code usage metrics, costs, and performance using OpenTelemetry, Prometheus, and Grafana.

Based on the [Sealos blog post](https://sealos.io/blog/claude-code-metrics).

Features

• Real-time Metrics Tracking: Monitor Claude Code sessions, token usage, costs, commits, and more
• Comprehensive Dashboard: 20+ panels showing key metrics, efficiency gauges, trends, and distributions
• Cost Analysis: Track costs by model, session, and time period
• Cache Efficiency Monitoring: Visualize prompt caching effectiveness
• Productivity Insights: Measure CLI vs user active time ratios
• Auto-provisioning: Dashboard and datasources automatically configured on startup

Architecture

Claude Code
    ↓ (OTLP HTTP/gRPC - metrics + logs)
OpenTelemetry Collector (:4318/:4317)
    ├─→ Prometheus metrics (:8889)
    │       ↓
    │   Prometheus (:9090)
    └─→ JSON logs (file)
            ↓
        Promtail
            ↓
        Loki (:3100)
            ↓ (PromQL + LogQL queries)
        Grafana (:3000)
            ↓ (Dashboard visualization)
        Your Browser

Metrics Tracked

• Sessions: Total Claude Code session starts
• Tokens: Usage by type (input, output, cache_read, cache_creation) and model
• Costs: Estimated USD costs with per-model breakdown
• Code Changes: Lines added/removed from accepted edits
• Git Activity: Commits and pull requests created
• Edit Decisions: Accept/reject choices on code suggestions
• Active Time: User input time vs CLI processing time
• Prompt Length: Average character count of user prompts (via Loki logs)

Prerequisites

• Docker and Docker Compose
• Claude Code CLI

Quick Start

1. Clone or Download This Repository

cd cc-metrics

2. Set Up Data Directories

make setup

Or manually:

mkdir -p data/prometheus data/grafana data/loki data/otel-logs
chmod