You just spent 6 hours pair-programming with Claude Code. You shipped three features, refactored a module, and wrote tests. But you have zero proof — no metrics, no record, nothing to show for it. Sound familiar?
AgentBoard fixes that. It's a free tool that auto-tracks your AI coding sessions and turns them into shareable stats — think Spotify Wrapped meets Strava, but for AI-assisted development. One command to install. No config. No behavior change needed.
TL;DR
AgentBoard auto-tracks your Claude Code (and soon Cursor) sessions, calculates metrics like AI Amplification and token consumption, generates beautiful shareable cards, and ranks you on a global leaderboard. Install with curl -sL agentboard.cc/install | bash. No account required to start. No source code ever leaves your machine.
How Does AgentBoard Work?
Three steps, 30 seconds total. AgentBoard reads your existing session metadata (never source code), uploads anonymized stats, and starts tracking automatically.
1. Run One Command
curl -sL agentboard.cc/install | bashThe CLI scans your local Claude Code session history, extracts aggregate metrics (tokens, tool calls, active time), and opens your personal stats card in the browser. No account needed — you can see your data immediately.
2. Sign In to Claim Your Profile
Connect GitHub, Google, or X to save your stats to a permanent profile at agentboard.cc/@yourhandle. Your terminal auto-links to your account — no second step, no API key to paste.
3. Code Normally
Every future AI coding session syncs in the background. You don't install plugins, run extra commands, or change anything about how you work. Just code.
What Metrics Does AgentBoard Track?
AgentBoard tracks seven core metrics that together paint a complete picture of your AI coding productivity:
- Active Time — How long you actively coded with AI assistance
- AI Equivalent Time — The estimated time your AI agent spent working (reading, editing, running commands)
- AI Amplification — The multiplier: AI time ÷ your active time. A 1.8x means you got 9 hours of output from 5 hours of work
- Tokens — Input and output token consumption, broken down by session
- Code Output — Lines added, deleted, and net change across all repos
- Messages & Tool Calls — How many prompts you sent and how many tools the AI used
- Tool Breakdown — Which tools (Bash, Read, Edit, Grep, etc.) your AI agent used, ranked by frequency
What About Privacy? What Don't You Track?
We never see your source code, conversation content, or file contents. AgentBoard only reads aggregate metadata — token counts, tool usage frequencies, timestamps, and line-change numbers.
Your code stays on your machine. We track how much you coded with AI, not what you coded. Full details in our methodology and privacy policy.
What Are Share Cards?
Share cards are beautiful, auto-generated images that summarize your daily AI coding stats. They come in multiple visual styles — minimal, wrapped, glassmorphism — and are designed to look great on X, LinkedIn, and Slack.
Think of them as your daily proof of work. Developers share them to flex their AI productivity, and it's become a ritual in some teams — posting your daily card like a Strava workout.
How Does the Leaderboard Work?
The global leaderboard ranks developers by daily AI coding activity — a composite of tokens consumed, code output, and AI Amplification. It updates in real time.
It's not just about total volume. Per-capita metrics mean a developer who codes efficiently for 3 hours can outrank someone who runs 8 hours of unfocused sessions. Quality of AI usage matters, not just quantity.
How Do I Get Started?
One command. Works with Claude Code today, with Cursor support coming soon.
curl -sL agentboard.cc/install | bashYou'll see your stats card in under 30 seconds. No credit card, no signup required to start. Sign in later to save your profile and join the leaderboard.