HN 표시: clude-telemetry – Claude Code용 다중 PC 사용량 대시보드

hackernews | | 📰 뉴스
#claude #claude code #cli #review #대시보드 #모니터링
원문 출처: hackernews · Genesis Park에서 요약 및 분석

요약

다수의 PC에서 Claude Code를 사용할 때 발생하는 기존 단일 머신 도구들의 비용 추적 한계를 극복하기 위해 'claude-telemetry'가 공개되었습니다. 이 시스템은 각 PC의 경량 에이전트가 ccusage를 기반으로 데이터를 수집하여 중앙의 Supabase 데이터베이스에 15분마다 자동 동기화하는 방식으로 작동합니다. 사용자는 다크 모드의 React 대시보드를 통해 머신, 프로젝트, 모델별로 통합된 비용을 분석하고, 5시간 단위의 사용량과 요금 제한 진행률 등을 직관적으로 확인할 수 있습니다. 또한 Cloudflare Workers 프록시를 통한 API 키 보호 및 승인된 이메일만 접근할 수 있는 화이트리스트 방식 등 강력한 보안 기능을 제공합니다.

본문

CLI tools like ccusage and ccost are single-machine. If you use Claude Code across multiple PCs, you have no unified view of your total spending. claude-telemetry solves this with an Elastic/Wazuh-style architecture: a lightweight Python agent on each PC auto-syncs usage data to a central Supabase database, and a React dashboard shows everything aggregated with filters by machine, project, model, and time period. The agent does no custom JSONL parsing — it calls ccusage as the parsing/pricing layer and focuses only on multi-PC aggregation and centralized sync. - Multi-PC aggregation with per-machine tracking - Auto-sync daemon (Elastic/Wazuh style agent) - Dark-mode dashboard with interactive charts (Recharts) - 5-hour block tracking with active block card, burn rate, and multi-PC timeline - Plan vs API cost comparison (Pro/Max 5x/Max 20x/Custom) - Rate limit progress bars (5-hour + weekly windows) - Usage pace calculator with trend detection - Project budget tracker with alerts at 90%/100% - Weekly usage reports with daily/weekly toggle - Cross-machine usage pace comparison - Model mix analysis (Opus/Sonnet/Haiku breakdown) - Supabase Auth (magic link login) with email whitelist - Cloudflare Workers proxy (zero exposed keys in frontend) - Deploy page with copy-paste agent install commands - Statusline auto-setup for rate limit tracking - Export data as CSV/JSON - Uses existing tools (ccusage) — no custom JSONL parsing - Create account at supabase.com - New Project — choose name and region - SQL Editor — paste and run each migration in order: - Authentication > URL Configuration: - Site URL: https://your-app.pages.dev - Redirect URLs: add same URL - Site URL: - Settings > API — copy Project URL and service_role key git clone https://github.com/RyanTech00/claude-telemetry.git cd claude-telemetry/dashboard npm install npx wrangler pages project create claude-telemetry npx wrangler pages secret put SUPABASE_URL # paste Project URL npx wrangler pages secret put SUPABASE_SERVICE_KEY # paste service_role key npx wrangler pages secret put ALLOWED_EMAILS # e.g. "[email protected],[email protected]" npm run build npx wrangler pages deploy dist cd claude-telemetry/agent python3 -m venv venv source venv/bin/activate # Windows: .\venv\Scripts\Activate pip install -e . claude-tracker setup claude-tracker sync --verbose claude-tracker install-service # auto-sync every 15min | Page | Description | |---|---| | Overview | Total cost, daily chart, model pie, machine cards, plan savings, rate limits | | Daily | Stacked area chart with daily/weekly toggle, top 10 days, hour heatmap | | Blocks | 5-hour block timeline, active block card, burn rate, multi-PC table | | Projects | Cost by project, pie distribution, budget progress bars, full table | | Models | Opus/Sonnet/Haiku breakdown, mix over time, savings alert | | Machines | Per-machine cards, comparison chart, daily pace, status badges | | Deploy | Generate agent install commands with one-click copy, statusline setup | | Sessions | Paginated table with sorting, filters by machine/project/type | | Insights | Plan value, burn rate, rate limits, budgets, trend analysis, cross-machine | | Settings | Plan selection, project budgets, alert thresholds, machine management, export | | Command | Description | |---|---| claude-tracker setup | Configure agent (interactive or --non-interactive ) | claude-tracker sync | Manual sync to Supabase | claude-tracker sync --verbose | Sync with detailed output | claude-tracker sync --force | Re-sync all data | claude-tracker daemon | Run auto-sync in foreground | claude-tracker daemon --interval 10 | Custom interval (minutes) | claude-tracker install-service | Install as system service | claude-tracker uninstall-service | Remove system service | claude-tracker service-status | Check daemon status | claude-tracker status | Show config and last sync | claude-tracker local --daily | View local data without syncing | claude-tracker setup-statusline | Configure rate limit tracking | claude-tracker uninstall | Remove agent config from this machine | - Dashboard access: Only emails listed in ALLOWED_EMAILS can log in. Set via:npx wrangler pages secret put ALLOWED_EMAILS # value: "[email protected]" or "[email protected],[email protected]" - API proxy: All Supabase keys are server-side only (Cloudflare Workers). Zero secrets in the browser bundle. - Agent auth: Each machine uses a unique API key stored locally in ~/.claude-tracker/config.json . - RLS: Row-Level Security ensures each user only sees their own machines and data. To completely remove claude-telemetry from all services: Step 1 — Remove agent (each PC) # Stop and remove the service claude-tracker uninstall-service # Remove config and data claude-tracker uninstall # Or manually: # Windows: rd /s /q %USERPROFILE%\.claude-tracker # Linux/macOS: rm -rf ~/.claude-tracker # Remove the package pip uninstall claude-usage-tracker # Delete the repo folder Step 2 — Delete Cloudflare Pages npx wrangler pages project

Genesis Park 편집팀이 AI를 활용하여 작성한 분석입니다. 원문은 출처 링크를 통해 확인할 수 있습니다.

공유

관련 저널 읽기

전체 보기 →