OpenAI Welcomes the Git AI Team Behind an Open-Source Agent Attribution Tool

OpenAI says it has welcomed Aidan Cunniffe and Sasha Varlamov from the Git AI team. In an approved post on X, OpenAI executive Tibo describes Git AI as an open-source tool that helps developers understand how coding agents contribute to their codebase and says the teams will work together to make that visibility more useful for businesses.

The announcement is not a new OpenAI-branded coding agent. The immediately actionable product is the existing Git AI open-source project, which describes itself as a Git extension for tracking AI-generated code. It records line-level attribution and links generated code to the agent, model, and prompts behind it.

Key Takeaways

• OpenAI says the Git AI team has joined the company.
• Git AI is open source and can be used locally without a login or team setup.
• Attribution is line-level: the project links code to the agent, model, and prompt session that generated it.
• git ai blame shows AI attribution alongside ordinary Git blame output.
• git ai stats reports aggregate AI-code, accepted-code, and tool/model breakdowns.
• Git AI uses Git Notes and is designed to preserve attribution through common history rewrites.
• The announcement does not establish an OpenAI acquisition, a renamed product, a hosted service launch, or an enterprise pricing change.

What Git AI Tracks

Git AI is designed to answer a question ordinary Git history does not: which parts of a file came from an AI coding session, and which agent and model produced them? The project says each attributed line can be linked to the agent, model, and prompts that generated it. (Git AI repository)

The project does not describe this as an AI detector. Instead, it records attribution reported by supported coding agents as they write code. That distinction matters: the tool is intended to preserve provenance from the agent session, not infer authorship later from the text alone. (Git AI repository)

Git AI stores attribution in Git Notes rather than changing the source files. Its documentation says the metadata follows common Git operations, including rebase, merge, stash, reset, cherry-pick, and branch changes, so teams can inspect provenance after normal repository maintenance. (Git AI standard; Git AI repository)

Inspecting Attribution From the Command Line

The project exposes two useful entry points:

git ai blame path/to/file
git ai stats --json

git ai blame is intended as a drop-in companion to git blame, showing which lines were attributed to a person or an AI session. git ai stats aggregates AI-generated lines, accepted lines, human overrides, and tool/model information across a repository or commit range. (Git AI repository)

This gives an engineering team two levels of visibility:

1. Line-level investigation: trace an implementation detail back to the agent session and prompt that produced it.
2. Repository- or range-level reporting: compare how much generated code was accepted, changed, or removed and which tools or models were involved.

The project also documents an attribution standard and support paths for adding more agents. Its README lists support for agents including Claude Code, Codex, Cursor, GitHub Copilot, OpenCode, Windsurf, and others, while the standard is intended to make the metadata interoperable across tools. (Git AI repository)

Installation And Operating Model

Git AI publishes install commands for macOS, Linux, and Windows. The project says the open-source CLI is local-first, works offline, requires no login, and does not require per-repository setup or a change to the normal commit workflow. (Git AI repository)

The project also says it avoids Git hooks and does not wrap the Git binary for its normal operation. That is a meaningful implementation choice for teams that want attribution without adding a slow hook to every repository or asking every developer to change how they commit. (Git AI repository)

The repository is Apache-2.0 licensed. Teams should still review the current installation instructions, supported-agent list, prompt-storage behavior, and the project’s documented limitations before deploying it across a company. (Git AI repository)

What OpenAI’s Announcement Does And Does Not Say

The approved OpenAI post says that Aidan Cunniffe and Sasha Varlamov from the Git AI team are joining OpenAI and that the collaboration will focus on helping businesses understand where coding agents contribute to their codebase. It does not say that OpenAI acquired Git AI, that the open-source repository will move, or that a new OpenAI-hosted Git AI service is available.

That boundary is important for adopters. Developers can evaluate and install the open-source Git AI CLI from its repository today, but the X announcement alone does not establish new OpenAI pricing, enterprise availability, support commitments, data-retention terms, or a timeline for future integration. (Tibo on X; Git AI repository)

Why Agent Attribution Matters

As coding agents take on larger changes, a commit-level label such as “AI-assisted” is often too coarse. Line-level attribution can help a team ask more precise questions:

• Which agent or model produced the code that reached production?
• How much generated code survived review and later edits?
• Which sessions or tools create repeated rework?
• Where did a human override an agent’s suggestion?
• Can a bug or incident be connected to the session that introduced the relevant code?

Git AI’s current CLI provides the attribution and statistics layer. The project separately describes a team offering for aggregate insights across repositories, but the OpenAI announcement does not establish any change to that offering or its access terms. (Git AI repository)

Bottom Line

OpenAI’s announcement puts the Git AI team behind an open-source code-attribution tool into the company’s orbit. For developers, the concrete tool is still the Git AI CLI: it records agent-reported provenance in Git Notes, exposes line-level git ai blame, and provides aggregate git ai stats without requiring a login or a new commit workflow. The future OpenAI integration story is not specified, so teams should treat the repository and its documentation—not the hiring announcement—as the source of current capabilities and setup instructions.

Sources

• Tibo on X, “Excited to welcome Aidan & Sasha from the Git AI team to OpenAI” (September 12, 2026): https://x.com/thsottiaux/status/2098569020815650816
• Git AI repository: https://github.com/git-ai-project/git-ai
• Git AI standard: https://github.com/git-ai-project/git-ai/blob/main/specs/git_ai_standard_v3.0.0.md