OpenAI has agreed to acquire Pierre Computer Company, a San Francisco developer-tools startup that builds Git-based storage for code written by AI agents. OpenAI announced the deal on November 4, 2026, describing it in its news feed as a way to expand the tools its Codex coding agent uses to build, understand and review code.
Neither company disclosed the price or how it will be paid. The deal has been signed but has not closed: Dealroom reported that it still needs regulatory approval.
Once the deal closes, OpenAI plans to build the features of Code Storage, Pierre's paid hosting service, into Codex, and to keep Pierre's open-source libraries available under an open-source license, SiliconANGLE reported. Over time the Pierre team is expected to help OpenAI ship new products and platforms, with a focus on software that makes coding agents faster and on agent workflows built around Git.
What Pierre built
Pierre describes itself as a maker of low-level, composable product infrastructure for machine work. Its main commercial product, Code Storage, is a distributed storage system built on Git that lets software, rather than people, create, read and update repositories at very high volume. SiliconANGLE noted that customers can also use it to keep related files such as software licenses and interface design assets.
According to Pierre's product site, the system uses Git's three-phase commit protocol and a quorum-based design to keep reads and writes durable and consistent. Developers reach it through SDKs for TypeScript, Python and Go or over the standard Git protocol, and a product can expose its own clone, push and fetch addresses under its own domain. The service also offers webhooks and a sync engine for GitHub, GitLab and Bitbucket.
Pierre advertises no caps on total storage or the number of repositories, a 32-terabyte limit per repository, support for Git Large File Storage and a 99.99% uptime commitment. Pricing is based on usage, with separate rates for frequently and rarely accessed data, and larger customers can run a managed version in their own cloud accounts.
Open-source tools already inside Codex
Alongside Code Storage, Pierre publishes several open-source projects. Diffs renders code changes and highlights the lines that were altered, either side by side or in line. Trees is a library for the file and folder sidebars found in code editors, with search, drag-and-drop and markers on recently changed files. The company also runs DiffsHub, and in September 2026 it released Highlights, a syntax highlighter built with WebAssembly that supports more than 70 languages.
Trees and Diffs already power file browsing and code review in the OpenAI Codex app, according to both Dealroom and SiliconANGLE, so OpenAI is buying a company whose software it already ships. AI Weekly reported that Diffs, Trees and DiffsHub will stay open source after the deal.
Customers, founders and funding
Pierre's site lists Lovable, Poke, Anything, Bolt, Amp and Parahelp among the companies that store generated code and other product files on Code Storage. In an August 2026 case study, Pierre said Lovable creates more than 8 million new repositories a week on the service, with peaks of 250 new repositories and 15,000 operations per second. Pierre put the improvement in Lovable's median fetch latency at 230%; SiliconANGLE described the gain as a cut of about 70% in code retrieval latency.
Neither OpenAI nor Pierre gave details of the founding team or funding in the sources available. According to Fast AI Jobs, a startup jobs database, Pierre was founded in 2023 by Jacob Thornton, a co-creator of the Bootstrap web framework, and Ian Ownbey, an early Twitter engineer. Thornton writes the case studies on Pierre's site.
The same database lists $23.5 million in total disclosed funding for Pierre, including a $23 million Series A in November 2025, and a team of between one and ten people. It names no investors, and these figures could not be confirmed with the company or OpenAI.
Why OpenAI is buying
The deal came two days after OpenAI described Pando, an engine for its internal Git environment that speeds up how its engineers' coding agents create and read code files. AI Weekly, citing OpenAI, reported that Pando cut median code-push latency by nearly 80%, moves close to a petabyte of Git data a day and went into production in about ten weeks, built with Codex agents.
Pierre gives OpenAI a commercial version of the same kind of infrastructure, tuned for agents that open and change repositories far faster than human developers. Several outlets, including SiliconANGLE and Dealroom citing Axios, framed the purchase as a move against GitHub, the code hosting service owned by Microsoft. AI Weekly noted that Microsoft also owns nearly a quarter of OpenAI.
Pierre is the latest developer-tools company OpenAI has agreed to buy this year. In June 2026 it announced plans to acquire Ona, which runs coding agents in cloud sandboxes.
A busy day for coding-agent deals
The Pierre agreement was one of three coding-agent deals announced on November 4. Meta acquired Conductor, whose app lets engineers run several coding agents at once, and Factory, which builds enterprise coding agents, acquired HumanLayer, a startup focused on team collaboration around agents.
Each of the three buyers is assembling tools around coding agents rather than the underlying models: storage, review and navigation in OpenAI's case, and orchestration and collaboration for Meta and Factory. The companies did not say what will happen to Code Storage's existing customers once the service moves into Codex.
