Skip to main content
autter blame shows the commit author and the coding agent responsible for each attributed line.

Inspect a file

The output follows git blame conventions and adds AI authorship where available. Standard git blame flags remain available.
Placeholder for Autter AI blame showing human and Codex-authored lines

Replace this placeholder with final terminal or editor attribution output.

Common examples

From a line to the prompt that wrote it

On a terminal, autter blame ends with a legend of the prompts behind its AI lines:
Follow one line to its session, model and change:
  1. autter blame <file> gives each prompt a short reference such as s_3f9a2c1d. Add --show-prompt to print the reference on every AI line; the legend is then printed even when output is piped.
  2. autter blame --why <file>:<line> explains one line: agent, model, prompt reference, the agent’s own session ID, the commit, and the prompt transcript.
  3. autter show-prompt <reference> prints the full prompt record as JSON, including the commit whose authorship note holds it.
  4. git show <commit> shows the change itself.
autter show-prompt accepts: Run autter show-prompt --help for the same list. Transcripts are read from your machine first, then from Autter when you are connected and signed in. Secrets are redacted before a transcript is stored. When a transcript can’t be shown, --why and show-prompt (in transcript.reason) say why: no transcript was recorded, it stayed on the machine that made the commit (local mode), or you need to run autter login. Authorship notes made on other machines arrive with autter fetch-notes.

Use blame during review

AI blame helps you answer:
  • Which agent session introduced this line?
  • Did a human edit the generated code before commit?
  • Which model produced the surrounding implementation?
  • Is the same session responsible for related lines in other files?
Use prompt access according to your organization’s permissions. Prompt content may be local-only, connected but restricted, or unavailable.

Editor support

The Autter VS Code extension can display color-coded AI attribution in the gutter for VS Code and compatible editors such as Cursor, Windsurf, and Antigravity.

Install the VS Code extension

View AI attribution beside the code you are reviewing.

JSON for tools

Use --json when building an editor, review tool, or internal analysis pipeline. Treat prompt and identity fields as sensitive data.

Data and privacy

Understand where attribution identities and prompt records are stored.