live-report

qmoni

Live session telemetry and Grafana dashboards for Claude Code via OpenTelemetry + JSONL transcript parsing.

What It Does

• Real-time monitoring — Configures Claude Code's native OpenTelemetry export to send metrics, events, and traces to a Grafana stack
• Pre-built dashboards — Ships Grafana dashboards for token usage, cost tracking, cache efficiency, tool usage, and more
• Historical analysis — Parses local JSONL session transcripts for per-session and cross-session analysis
• Auto-configuration — SessionStart hook automatically sets OTEL env vars when configured

Architecture

Claude Code → OTLP gRPC :4317 → OTel Collector → Prometheus (metrics)
                                                → Loki (events/logs)
                                                → Tempo (traces)
                                                → Grafana (dashboards)

Installation

Install via Claude Code Marketplace (Recommended)

Step 1 — Add the marketplace:

# From inside Claude Code
/plugin marketplace add phuquocchamp/qmoni

# Or from the CLI
claude plugin marketplace add phuquocchamp/qmoni

Step 2 — Install the plugin:

# From inside Claude Code
/plugin install live-report@qmoni

# Or from the CLI
claude plugin install live-report@qmoni

Install from local clone

git clone https://github.com/phuquocchamp/qmoni.git
claude plugin marketplace add ./qmoni
/plugin install live-report@qmoni

Verify installation

claude plugin list

You should see live-report in the output.

Quick Start

1. Install the plugin (see above)

2. Run setup inside Claude Code:

   /live-report setup
   

3. Restart Claude Code. Telemetry begins automatically.

4. Open dashboard:

   /live-report dashboard
   

Commands

Command	Description
/live-report setup	Configure OTEL env vars + deploy Docker stack
/live-report start	Verify telemetry is active and collector reachable
/live-report status	Show stack health and data freshness
/live-report dashboard	Open Grafana in browser
/live-report history	Parse JSONL transcripts for historical analysis

Configuration

Create .claude/live-report.local.md in your project (created automatically by setup):

---
enabled: true
grafana_url: http://localhost:3000
otlp_endpoint: http://localhost:4317
otlp_protocol: grpc
traces_beta: true
---

Prerequisites

• Docker and Docker Compose (for the monitoring stack)
• Claude Code v2.1+ (for OpenTelemetry support)
• jq (for JSONL parsing in history command)

Docker Stack

The included stack runs 5 services:

Service	Port	Purpose
OTel Collector	4317, 4318	Receives OTLP data
Prometheus	9090	Stores metrics
Loki	3100	Stores events/logs
Tempo	3200	Stores traces
Grafana	3000	Dashboards

Manage manually:

# Start
docker compose -f stack/docker-compose.yml up -d

# Stop
docker compose -f stack/docker-compose.yml down

# Logs
docker compose -f stack/docker-compose.yml logs -f

Dashboard Panels

1. Token Usage — Stacked bar chart (input, output, cacheRead, cacheCreation)
2. Session Cost — Running total with color thresholds
3. Cost Over Time — Rate of spend
4. Active Sessions — Counter
5. Active Time — User vs CLI time
6. Lines of Code / Commits / PRs — Productivity stats
7. Cache Efficiency — Gauge (cache hit ratio)
8. Token Usage by Model — Pie chart
9. Tool Usage — Table from Loki events
10. API Latency — Histogram

Privacy

All telemetry is local (localhost). No data leaves your machine unless you configure an external endpoint. Prompt content, tool details, and API bodies are disabled by default and require explicit opt-in via environment variables.

License

MIT