AI coding agents compared across 402 repos
Each of the 402 repos on the board has a score for every agent. Each comparison puts two side by side: averages, files each reads, checks each weighs, and repos that favour one.
The scores measure how well a repo is set up for an agent, not which agent writes better code.
Which pair do you want to compare?
| Comparison | Averages | Higher for each |
|---|---|---|
| Claude Code vs Cursor | 71.9 vs 63.5 | 388 vs 12 |
| Claude Code vs Codex | 71.9 vs 70.5 | 287 vs 107 |
| Gemini CLI vs Claude Code | 63.5 vs 71.9 | 4 vs 396 |
| Cursor vs Codex | 63.5 vs 70.5 | 14 vs 388 |
| Aider vs Claude Code | 63.3 vs 71.9 | 44 vs 358 |
| OpenHands vs Claude Code | 64.5 vs 71.9 | 15 vs 384 |
| Pi vs Claude Code | 68.5 vs 71.9 | 51 vs 348 |
| Kimi Code CLI vs Claude Code | 69.2 vs 71.9 | 71 vs 331 |
| Devin vs Claude Code | 68.0 vs 71.9 | 36 vs 361 |
| Codex vs Gemini CLI | 70.5 vs 63.5 | 397 vs 5 |
“Higher for each” counts the repos that score higher for the first agent, then the second; the rest tie.