Skip to content
sobolewski.proAI today, every eveningSubscribe

Once a day. Unsubscribe with one click. Privacy policy

OpenAI acquires Pierre: Git for AI agents is coming to Codex

OpenAI acquires Pierre Computer, the makers of Code Storage, a Git built for AI agents, and will build it into Codex. Here is what that changes.

3 min read

This article in other languages:العربيةDeutschEspañolFrançaisItaliano日本語Polski简体中文

In this article
  1. What exactly is OpenAI buying?
  2. Why is OpenAI acquiring Pierre an interesting move?
  3. Is this a challenge to GitHub?
  4. What should we watch?
  5. FAQ

OpenAI announced on November 4 that it will acquire Pierre Computer Company, a developer tools startup. According to OpenAI, acquiring Pierre will expand Codex’s tools for writing, understanding and reviewing code. Axios broke the story first; no price was disclosed.

What exactly is OpenAI buying?

Pierre’s main product is Code Storage, a distributed Git-based repository service with an API designed for machines rather than for people clicking through a web interface. Apps and AI agents use it to create repositories, write code and track its history through the regular Git protocol (clone, push, fetch) or through an SDK.

According to the product page, Code Storage supports, among other things:

  1. repositories of up to 32 TB and more than 30 concurrent writes,
  2. Git LFS, webhooks and SDKs for TypeScript, Python and Go,
  3. sync with GitHub, GitLab and Bitbucket.

Code Storage is already used by Lovable, Bolt, Amp, Poke, Anything and Parahelp, among others, which covers a good share of the tools where AI generates entire apps. Interestingly, Pierre’s technology already runs inside Codex, and after the acquisition it will be integrated more deeply.

Pierre also builds open source projects: Diffs, a library for displaying code changes, plus Trees and DiffsHub. According to reports, they will stay open source after the acquisition.

Why is OpenAI acquiring Pierre an interesting move?

Because AI agents write code at a pace Git and GitHub were never designed for. A single developer with a few AI agents running in parallel creates more branches, pushes and merges than a whole team did a few years ago.

OpenAI knows this problem firsthand. Two days earlier, on November 2, the company described Pando, its own Git-compatible backend built for the scale Codex makes possible, in the post “Scaling Git at OpenAI”. According to OpenAI, after the move to Pando median push time dropped by almost 80%, and the system moves close to a petabyte of Git data per day (we know these numbers from coverage of the post).

So Pierre is the second piece of the same puzzle: Pando runs inside OpenAI, while Code Storage is a finished product that outside companies already use.

Is this a challenge to GitHub?

That is how the media describe the deal, including Axios, whose headline links it to the growing pressure AI coding puts on GitHub. The twist is that GitHub belongs to Microsoft, and Microsoft is one of OpenAI’s largest shareholders.

For now, though, this is more a direction than a ready-made competitor. OpenAI has not announced a public “GitHub for Codex” and has not said whether Pando and Code Storage will merge into one product. What is clear is that Codex is meant to be more than an agent that writes code: it should also store repositories, show changes and handle reviews. It is the same trend as the shared usage limit for ChatGPT and Codex, with OpenAI weaving chat, the agent and developer tools ever more tightly into a single service.

What should we watch?

First, the terms. We don’t know the price, and the deal still has to close, so until then the two companies operate separately.

Second, Code Storage customers. Lovable and Bolt build their products on infrastructure that is going to OpenAI. How that affects their plans is unknown for now (if we build something on Code Storage, it’s worth following Pierre’s announcements).

Third, open source. Diffs and Trees are supposed to stay open, but it’s worth seeing whether development slows down once the team moves to OpenAI.

For us as users, the upside is simple: faster, more reliable repositories for agents in Codex. The cost is even deeper dependence on a single vendor that owns the model, the agent, and now also the place where we keep our code.

FAQ

Who is acquiring Pierre Computer?

OpenAI. The company announced the acquisition on November 4, 2026 and plans to build Pierre’s technology, above all Code Storage, into Codex.

How much will OpenAI pay for Pierre?

Unknown. Neither OpenAI nor Axios disclosed the price.

What is Pierre’s Code Storage?

It’s a distributed, Git-based code storage system with an API for apps and AI agents. It supports repositories of up to 32 TB, Git LFS, and sync with GitHub, GitLab and Bitbucket.

Will Pierre’s Diffs and Trees stay open source?

According to reports on the acquisition, yes: Diffs, Trees and DiffsHub are set to remain open source projects.

  • AI
  • OpenAI
  • AI agents
  • programming

AI today, every evening

The day's biggest AI news and new posts from the blog, every evening. Short, with sources.

Once a day. Unsubscribe with one click. Privacy policy