Claude Code Metrics Setup: Prometheus + Grafana dashboards for tracking AI coding productivity

This guide walks through setting up OTEL telemetry from Claude Code to Prometheus + Grafana.

Architecture #

Claude Code  -->  OTLP/HTTP  -->  Prometheus  -->  Grafana
   (metrics)      (protobuf)      (storage)       (viz)

1. Environment Variables #

Add to your shell profile (~/.bashrc for Linux, ~/.zshrc for macOS):

export CLAUDE_CODE_ENABLE_TELEMETRY=1

export OTEL_METRICS_EXPORTER=otlp
export OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
export OTEL_EXPORTER_OTLP_ENDPOINT=http://localhost:9090/api/v1/otlp
export OTEL_EXPORTER_OTLP_METRICS_TEMPORALITY_PREFERENCE=delta

Reload your shell: source ~/.bashrc

2. Prometheus Setup #

Install (AL2023 / Amazon Linux)

cd /tmp
curl -LO https://github.com/prometheus/prometheus/releases/download/v3.0.0/prometheus-3.0.0.linux-amd64.tar.gz
tar xzf prometheus-3.0.0.linux-amd64.tar.gz
sudo mv prometheus-3.0.0.linux-amd64 /opt/prometheus

mkdir -p ~/.prometheus/data

Install (macOS with Homebrew)

brew install prometheus

Configuration

Create ~/.prometheus/prometheus.yml:

global:
  scrape_interval: 15s

scrape_configs:
  - job_name: "prometheus"
    static_configs:
    - targets: ["localhost:9090"]

Run Prometheus with OTLP Receiver

Critical flags:

• --web.enable-otlp-receiver - Enables the OTLP ingestion endpoint
• --enable-feature=otlp-deltatocumulative - Converts delta metrics (like lines_of_code) to cumulative

/opt/prometheus/prometheus \
  --config.file=~/.prometheus/prometheus.yml \
  --storage.tsdb.path=~/.prometheus/data \
  --web.listen-address=127.0.0.1:9090 \
  --web.enable-otlp-receiver \
  --enable-feature=otlp-deltatocumulative

prometheus \
  --config.file=/opt/homebrew/etc/prometheus.yml \
  --storage.tsdb.path=/opt/homebrew/var/prometheus \
  --web.listen-address=127.0.0.1:9090 \
  --web.enable-otlp-receiver \
  --enable-feature=otlp-deltatocumulative

You can run this in a tmux session or background it.

3. Grafana Setup #

Install (AL2023 / Amazon Linux)

sudo tee /etc/yum.repos.d/grafana.repo << 'EOF'
[grafana]
name=grafana
baseurl=https://rpm.grafana.com
repo_gpgcheck=1
enabled=1
gpgcheck=1
gpgkey=https://rpm.grafana.com/gpg.key
sslverify=1
sslcacert=/etc/pki/tls/certs/ca-bundle.crt
EOF

sudo dnf install grafana -y
sudo systemctl start grafana-server

Install (macOS)

brew install grafana
brew services start grafana

Access

• URL: http://localhost:3000
• Default login: admin / admin

Add Prometheus Data Source

1. Go to Connections → Data Sources → Add data source
2. Select Prometheus
3. URL: http://localhost:9090
4. Save & Test

Import Dashboards

1. Go to Dashboards → Import
2. Paste the JSON from the dashboard files in this directory
3. Select your Prometheus data source
4. Import

Dashboard files included:

• dashboard-metrics.json - Claude Code Metrics (tokens, cost, productivity ratio)
• dashboard-summary.json - Daily/Weekly Summary (aggregated stats)
• dashboard-economics.json - Engineering Economics (ROI, team equivalence)

4. Verify Metrics Are Flowing #

After using Claude Code with telemetry enabled:

curl -s 'http://localhost:9090/api/v1/label/__name__/values' | jq '.data[]' | grep claude

You should see:

• claude_code_token_usage_tokens_total
• claude_code_cost_usage_USD_total
• claude_code_active_time_seconds_total
• claude_code_lines_of_code_count_total
• claude_code_session_count_total

5. Available Metrics #

Metric	Labels	Description
claude_code_token_usage_tokens_total	type (input/output/cacheRead/cacheCreation), model	Token consumption
claude_code_cost_usage_USD_total	model	Estimated API cost in USD
claude_code_active_time_seconds_total	type (cli/user)	Time tracking
claude_code_lines_of_code_count_total	type (added/removed)	Code output (delta metric!)
claude_code_session_count_total		Session count
claude_code_event_count_total	event	Events (commits, PRs, etc.)

6. Key Queries #

Productivity Ratio (CLI time / User time):

sum(claude_code_active_time_seconds_total{type="cli"})
/ sum(claude_code_active_time_seconds_total{type="user"})

Lines per Dollar:

sum(sum_over_time(claude_code_lines_of_code_count_total[$__range]))
/ sum(increase(claude_code_cost_usage_USD_total[$__range]))

Cache Efficiency:

sum(claude_code_token_usage_tokens_total{type="cacheRead"})
/ (sum(claude_code_token_usage_tokens_total{type="cacheRead"}) + sum(claude_code_token_usage_tokens_total{type="input"}))
* 100

Important: The lines_of_code metric is a delta metric, not cumulative. Use sum_over_time() instead of increase():

sum(sum_over_time(claude_code_lines_of_code_count_total[$__range]))

sum(increase(claude_code_lines_of_code_count_total[$__range]))

7. Troubleshooting #

No metrics showing up

1. Check env vars are set: env | grep OTEL
2. Verify Prometheus is running with OTLP flags: ps aux | grep prometheus
3. Check Prometheus logs for OTLP errors
4. Make sure you've used Claude Code after setting env vars (restart terminal first)

Grafana can't connect to Prometheus

1. Verify Prometheus is running: curl http://localhost:9090/-/healthy
2. Check Grafana data source URL matches Prometheus listen address

Dashboard shows no data

1. Check time range (try "Last 24 hours")
2. Verify metrics exist: curl 'http://localhost:9090/api/v1/query?query=claude_code_token_usage_tokens_total'
3. Make sure Prometheus data source is selected in dashboard

8. For Datadog (Alternative) #

If pushing to Datadog instead of local Prometheus:

export OTEL_METRICS_EXPORTER=otlp
export OTEL_EXPORTER_OTLP_PROTOCOL=http/protobuf
export OTEL_EXPORTER_OTLP_ENDPOINT=https://otel.datadoghq.com
export OTEL_EXPORTER_OTLP_HEADERS="DD-API-KEY=your_api_key"

The same metrics will flow to Datadog, and you can build similar dashboards there.

source & further reading

gist.github.com — original article agent-proxy — a zero-dep proxy that shows the bloat in Claude Code's requests (ranked tool table + full readable Markdown of every request) PowerShell profile and test suite A plan-of-plans for a production-grade end-to-end e-commerce website.