Know what AI wrote — and why

AI Blame extends git blame with AI authorship. See which agent and model wrote each line, and the prompt that generated it.

git-ai blame src/config.ts
12
claude-code(opus)
export async function loadConfig(path: string) {
13
cursor(gpt-4o)
const raw = await fs.readFile(path, 'utf-8');
14
cursor(gpt-4o)
const parsed = JSON.parse(raw);
15
human
// validate required fields
16
claude-code(sonnet)
if (!parsed.version) {
17
claude-code(sonnet)
throw new ConfigError('missing version');
18
claude-code(sonnet)
}
19
copilot(gpt-4o)
return parsed as AppConfig;

Agent Attribution

Cross-agent attribution for every line

Know exactly which agent and model wrote every line of code — across Claude Code, Copilot, Cursor, and any other tool. AI Blame works with any AI coding agent, giving you a unified view of authorship.

  • See which agent and model authored every line in any file
  • Works with Claude Code, Copilot, Cursor, Windsurf, and more
  • Color-coded annotations in your IDE gutter or terminal
Book a Demo

Prompt Context

See the prompt behind every line

Hover in your IDE or run `git-ai blame` in the terminal to see the exact prompt that generated any line of code. Understand not just who wrote it, but why — the intent and context behind every AI-generated change.

  • Run `git-ai blame` in the terminal for full prompt context
  • Hover annotations in your IDE to see prompts inline
  • Link from any line back to the full conversation that created it
Book a Demo

Open Standard

Built on an open standard

Your AI attribution data is portable and version-controlled. Never locked into a specific agent or vendor — AI Blame uses an open format that any tool can read and write.

  • Open format — no vendor lock-in for your attribution data
  • Data lives alongside your Git history, fully version-controlled
  • Any agent or tool can read and write AI attribution metadata
Book a Demo

AI Blame

Know what AI wrote — and the prompt behind it

Extend git blame with AI authorship. See which agent wrote every line, trace it back to the prompt, and maintain full context as your AI-generated codebase grows.