Agent Blame

Agent Blame shows you which lines of code in a GitHub Pull Request were written by AI, so you can focus your code reviews where it matters.

How It Works

• Orange gutter marks highlight AI-generated lines
• Hover over marks to see prompt details (agent, model, and prompt text)
• Unmarked lines are human-written (clean, distraction-free UI)
• File badges show AI percentage per file
• PR summary shows total AI vs human code stats

• Full analytics dashboard in GitHub Insights → Agent Blame
• Track AI vs human code trends over time
• See tool breakdown (Cursor, Claude Code, OpenCode)
• Model usage statistics
• Per-contributor AI usage
• Commit:Prompt efficiency ratio
• Filter by time period (24h, 3 days, week, month, all time)

• Visual markers directly on GitHub PR diff views
• Works in both unified and split diff modes
• Supports light and dark themes
• Zero performance impact on page load

• A GitHub Personal Access Token (with repo scope)
• The https://github.com/mesa-dot-dev/agentblame installed in your repositories to capture AI attribution

1. Install the extension
2. Click the extension icon
3. Add your GitHub token
4. Visit any PR with agentblame attribution data
5. Check Insights → Agent Blame for repository analytics

• Your GitHub token is stored locally, never sent to external servers
• No analytics, tracking, or data collection
• Open source: https://github.com/mesa-dot-dev/agentblame