Commit Stats

The stats command gives you basic stats for a single commit, or a sequence of linear commits. It is not accurate for complicated or cyclic histories, and not performant for large ranges. To compute repo-level or PR-level AI percentages, use the SCM Bot built for teams.

Single Commit

Show stats for the current HEAD or a specific commit:

git-ai stats
git-ai stats <commit-sha>

Commit Ranges

Pass a range to aggregate stats across multiple commits:

git-ai stats <start>..<end>

To compute stats for the entire repository history, use the empty tree SHA as the start:

git-ai stats 4b825dc642cb6eb9a060e54bf8d69288fbee4904..HEAD

Stats are computed per-commit from git notes, so large ranges take proportionally longer. For repositories with thousands of commits, Git AI for Teams precomputes these aggregations and serves them from a dashboard.

JSON Output

Add --json to get machine-readable output:

git-ai stats --json
git-ai stats <commit-sha> --json
git-ai stats <start>..<end> --json

Example response:

{
  "human_additions": 12,
  "mixed_additions": 6,
  "ai_additions": 87,
  "ai_accepted": 81,
  "total_ai_additions": 112,
  "total_ai_deletions": 23,
  "time_waiting_for_ai": 42,
  "git_diff_added_lines": 93,
  "git_diff_deleted_lines": 14,
  "tool_model_breakdown": {
    "cursor/claude-4.5-opus": {
      "ai_additions": 81,
      "mixed_additions": 4,
      "ai_accepted": 77,
      "total_ai_additions": 98,
      "total_ai_deletions": 18,
      "time_waiting_for_ai": 38
    },
    "claude-code/claude-opus-4-5-20251101": {
      "ai_additions": 6,
      "mixed_additions": 2,
      "ai_accepted": 4,
      "total_ai_additions": 14,
      "total_ai_deletions": 5,
      "time_waiting_for_ai": 4
    }
  }
}

Field descriptions:

Lines in the commit:

• human_additions — Lines attributed to humans, including AI lines edited before commit.
• ai_additions — Lines attributed to AI, including lines later edited by humans.
• mixed_additions — AI-generated lines edited by a human before commit.
• ai_accepted — AI-generated lines committed without any human edits.

Total AI activity:

• total_ai_additions — All lines generated by AI during the working session, whether or not they appear in the final diff.
• total_ai_deletions — All lines deleted by AI during the working session, whether or not those deletions remain in the final diff.

Timing:

• time_waiting_for_ai — Wall-clock seconds spent waiting for AI responses.

Git baseline:

• git_diff_added_lines — Added lines reported by the raw git diff.
• git_diff_deleted_lines — Deleted lines reported by the raw git diff.

Per-tool:

• tool_model_breakdown — Object keyed by <tool>/<model> with the same metrics scoped to that tool and model pair.

human_additions + ai_additions can exceed git_diff_added_lines because mixed_additions count toward both.

Ignoring Files

Stats automatically exclude common non-source artifacts: lockfiles, generated/minified files, vendor dependencies, and snapshot artifacts. Files marked linguist-generated in .gitattributes are also excluded.

Add custom ignore patterns with --ignore:

# Exact filename
git-ai stats --ignore Cargo.lock

# Glob pattern (quote to prevent shell expansion)
git-ai stats --ignore "*.lock"

# Multiple patterns
git-ai stats --ignore Cargo.lock "*.generated.*" package.lock

For persistent ignore rules, create a .git-ai-ignore file in the repository root:

# Comments allowed
docs/generated/**
*.snap
vendor/**

Effective ignore patterns are the union of built-in defaults, .gitattributes (linguist-generated), .git-ai-ignore, and CLI --ignore flags. --ignore is additive — it does not replace defaults.

Beyond Commits: Pull Requests, Cost, and Prompt Data

git-ai stats operates on commit-level data stored in git notes. This covers authorship attribution, line counts, and timing — everything recorded locally during a coding session.

Some metrics require processing that goes beyond what git notes contain. Joining attributions on SCM metadata, computing per-PR metrics while handling squash merges and rebases, tracking token costs across models, and analyzing prompt effectiveness all depend on additional telemetry and server-side computation.

Git AI for Teams provides precomputed dashboards with per-PR metrics, model cost tracking, and prompt analysis. For teams that want to join the data with other sources, Git AI supports exporting to data warehouses.