AI Code Attribution · Blamely

Trace AI-written code. Understand its impact.

Your code. Your attribution. Your Git. Blamely runs locally, including in air-gapped environments. Connect code origin to review, rework, and delivery with Oobeya.

Get started with Blamely (opens in a new tab)Schedule a demo
Blamely attribution + Oobeya engineering contextFrom code origin to delivery outcomes

Performance Overview

AI attribution alongside delivery and usage

AI share of new code61%↑ +3.7%

75,403 of 123,666 added lines are AI-attributed.

AI impact on deliveryExplore the trend
Delivery velocity

158commits/week

↓ −2%prev. 161.3
Time per commitderived

36.2min

↓ −53.3%prev. 77.6 min
Reclaimed capacityderived

1,401hours

2,031 commits × 41.4 min saved
Developers using AIderived

17developers

↑ +54.5%65.4% of 26 developers
Coding time split
219.4h AI40.8h Human
Avg. lines / AI commit
114.4 AI83.5 Human
Token spend
905.5M
Cache efficiency
99.3%

Explore AI contribution, then compare it with delivery and quality trends.

Attribution across the assistants your team uses

  • Claude Code
  • Cursor
  • GitHub Copilot
  • Codex CLI
  • Gemini CLI
  • Devin
  • OpenCode

macOS, Windows, and Linux · CLI and editor integrations · Capture varies by tool

How it works

Code origin stays connected to the change

Start with the source of a change. Add the engineering context needed to decide what to investigate.

Read the technical documentation (opens in a new tab)

  1. 1. Capture edits locally

    A background CLI records supported assistant edits. Editor extensions add signals such as accepted inline completions.

  2. 2. Attribute each commit

    Changed lines are matched to their source and stored under refs/notes/blamely. Push those notes when you choose.

  3. 3. Evaluate the engineering impact

    Use Oobeya to investigate AI adoption alongside review time, rework, quality, and delivery performance.

In the developer workflow

See the source while reading the code

Blue and orange gutter marks distinguish AI and human contributions. Inspect provenance, file changes, and the current attribution split inside the editor.

Local capture. Attribution stored in Git.
Explore editor setup (opens in a new tab)
Blamely editor illustration with AI and human contribution percentages, file changes, and per-line gutter marks
Editor illustration from Blamely.

Commit-level evidence

Inspect the contribution behind a commit

Open the example report (opens in a new tab)

Generate a local HTML report with blamely report --html. Inspect AI share, tools, models, and attributed file ranges.

Example Blamely commit report showing AI and human attribution, generation types, contributing tools, file ranges, and tokens
Commit report illustration from Blamely. Example values are shown.

Blamely + Oobeya

Go beyond the percentage of AI-written code

A higher AI share can coincide with faster delivery, more review work, or more rework. Compare the signals before deciding what to change.

Engineering questions and product context for AI attribution metrics
MetricQuestion to investigateProduct context
AI code shareWhat portion of work is AI-assisted or AI-generated?Blamely attribution reports
Pull request cycle timeHow do AI-assisted changes move through review?AI Impact + PR analytics
Code churnIs AI-generated code being rewritten, reverted, or deleted?Attribution + code churn
Review loadWhere is review effort increasing?Review analytics
Quality trendsHow does AI-assisted work relate to defects and quality findings?Quality + delivery metrics
Cost contextHow does AI spending relate to adoption and delivery trends?AI Impact + cost monitoring

Explore further

Build a fuller view of AI-assisted development

AI Impact

Evaluate AI coding assistant adoption alongside delivery and quality outcomes.

Explore AI Impact

AI Insights

Review strengths, risks, and areas for improvement across AI-assisted engineering.

Explore AI Insights

Common questions

AI code attribution questions

Need to discuss your team or technical setup? Contact us.

What does AI code attribution measure?

It records whether committed lines came from a person or a supported AI assistant. Blamely also identifies the assistant and model where that information is available.

How is this different from git blame?

Git blame identifies the commit author. Blamely captures the origin of the code, including AI suggestions accepted by that author.

Does Blamely upload my source code?

The CLI, daemon, and attribution database run locally. Attribution is stored in Git notes. Connecting those signals to Oobeya is a separate integration step.

Which assistants and editors are supported?

Claude Code, Cursor, GitHub Copilot, Codex CLI, Gemini CLI, Devin, and OpenCode. Editor integrations include VS Code, Cursor, JetBrains, and Antigravity. Capture methods differ by tool.

Can it identify AI code written before installation?

Attribution relies on captured tool activity and available attribution records. Git history alone does not reliably reveal whether a past change was written by AI.

Does higher AI code share mean higher productivity?

AI share describes how code was produced. Evaluate it alongside review effort, code churn, quality, delivery, and cost to understand the results. A correlation alone does not establish causation.

See Oobeya in practice

Connect code origin to engineering outcomes

See how Blamely attribution and Oobeya help your team investigate AI adoption, review effort, and delivery performance.

Schedule a demoContact us