commands.rs×
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
This is here to make sure we can handle stdin across every supported shell
Opus 4.5 · John Smith
Help developers understand large AI-generated codebases
The “why” for every line. Git AI links AI-generated code to the prompts that created it. Intent, requirements, and decisions stay with the code—forever.
- Full prompt history for every AI-generated line of code
- Preserve the "why" behind every line of AI-Code
- Ask the agent who wrote any section's code questions about it
Built into the IDE:VSCodeCursorWindsurfClineEmacsIntelliJsoon
AI Blame
See which code was AI-generated and what it's supposed to do, in the IDE.
- Git-native blame annotations for AI-generated code
- IDE integration: hover any line to see its AI origin
Cloud
Git AI managed
Prompt Store
Self-Hosted
Your infrastructure
Prompt Store
auth.ts
1
human
2
human
3
ai
4
ai
5
ai
6
human
7
human
Own your prompts
Store full agent transcripts or summaries. Git AI makes prompts portable and lets you store them on your own infrastructure.
- All prompt data stays within your infrastructure
- Portable format — never locked into any vendor's platform
- Built-in access controls to use prompts as context
Agent Session · Plan Mode
✓Exploring codebase10 tools
✓Reading past prompts(/ask)
“We avoid the Sentry SDK here because we need to manage our own batching and flushing outside the CLI — otherwise it hangs before exiting. I'll keep the manual approach.”
Past Prompts Make Agents Smarter
When agents can read past prompts and understand what existing code is supposed to do, they make fewer mistakes and produce more maintainable code.
- Agents plan better when they know the intent behind the code they're working with
- Smarter agents. More one-shots
- Agents respect past architecture decisions