Last reviewed
May 31, 2026
Sources
16 citations
Review status
Source-backed
Revision
v2 · 2,044 words
Add missing citations, update stale details, or suggest a clearer explanation.
OpenAI's acquisition of Astral is a planned purchase, announced on March 19, 2026, in which OpenAI agreed to acquire Astral Software, the company behind a set of widely used Python developer tools written in Rust. Astral makes uv (a package and project manager), Ruff (a linter and formatter), and ty (a type checker). Under the agreement the Astral team, led by founder and chief executive Charlie Marsh, would join the team that builds OpenAI Codex, the company's AI coding system. OpenAI and Astral both said the open source tools would keep being supported and developed after the deal closes. Financial terms were not disclosed. [1][2][3]
The acquisition drew unusual attention because it was one of the first times a large AI company had bought a project that millions of developers already relied on for everyday work that has nothing to do with AI. That raised a question across the Python community about who controls foundational tooling, and what happens to it once a frontier lab owns the team that maintains it. [4][5]
Astral was founded in 2022 by Charlie Marsh, a developer based in Brooklyn, New York. The company set out to make the Python ecosystem faster and more pleasant to work in by rebuilding common Python tools in Rust, a systems language known for speed and memory safety. The pitch was simple: take tools that Python developers used every day, and make them run tens or hundreds of times faster. [6][7]
Its first widely adopted project was Ruff, a linter and code formatter. Ruff can stand in for Flake8 and many of its plugins, plus Black, isort, pydocstyle, pyupgrade, and several other tools, while running far faster than any of them individually. It supports hundreds of lint rules, many borrowed from the tools it replaces. Because it is a single fast binary, teams could fold a dozen separate checks into one step. [6][8]
The company's best known project is uv, a package and project manager released in 2024. uv is a drop-in replacement for pip, pip-tools, virtualenv, and parts of pipx and pyenv, rolled into one tool. It installs packages, builds virtual environments, locks dependencies, installs Python itself, and runs command line tools, all from a single binary that works on macOS, Linux, and Windows. Its speed comes from aggressive caching, parallel downloads, and a dependency resolver written in Rust. Astral's own benchmarks showed uv installing a set of packages from a warm cache in a fraction of a second where pip took several seconds. By the time of the acquisition, the company said its tools drew hundreds of millions of downloads a month across Ruff, uv, and ty. [1][9][10]
ty is the newest of the three. Announced in December 2025, it is a type checker and language server, again written in Rust, pitched as a faster alternative to mypy, Pyright, and Pylance. It had not reached the adoption of uv or Ruff by early 2026, but it followed the same performance-first approach, with Astral reporting it running many times faster than existing checkers. [11][12]
Astral had also started to build a commercial business on top of the free tools. In August 2025 it launched pyx, described as a Python-native package registry and the first piece of a paid Astral platform, with features aimed at hosting prebuilt application components including AI frameworks. The free tools were the on-ramp; pyx was meant to be the part that earned revenue. [9][13]
The table below lists Astral's main projects.
| Project | Type | Replaces or competes with | First released | Language |
|---|---|---|---|---|
| Ruff | Linter and formatter | Flake8, Black, isort, pydocstyle, pyupgrade | 2022 | Rust |
| uv | Package and project manager | pip, pip-tools, virtualenv, pipx, pyenv | 2024 | Rust |
| ty | Type checker and language server | mypy, Pyright, Pylance | 2025 (announced Dec 2025) | Rust |
| pyx | Package registry (commercial) | Private PyPI mirrors and registries | 2025 (beta) | n/a |
OpenAI announced the deal on March 19, 2026, through its newsroom and a post on the OpenAI site, and Astral published its own note titled "Astral to join OpenAI" the same day. In Astral's words, the company had "entered into an agreement to join OpenAI as part of the Codex team." [1][2]
OpenAI framed the purchase around Codex, its AI coding product. The stated goal was to move Codex past simply generating code and toward systems that can take part in the whole development workflow: planning changes, editing codebases, running tools, checking results, and maintaining software over time. Bringing in Astral's tools and engineers, OpenAI said, would speed up that work and let AI agents work more directly with the tools developers already use. Codex had grown quickly, with OpenAI reporting more than 2 million weekly active users and roughly a threefold increase in user growth since the start of the year. [1][3]
Charlie Marsh cast the move as a bet on where software development is heading. He wrote that AI was rapidly changing how software gets built and that the pace was only accelerating, and he described building at the intersection of AI and software as the highest-leverage thing the team could do. He also restated the company's long-running aim, "to make programming more productive," and "to build tools that radically change what it feels like to build software." [2][3]
The companies did not reveal the price. CNBC, Bloomberg, and other outlets reported the deal as announced rather than closed, and noted that it remained subject to customary closing conditions, including regulatory approval. Until closing, OpenAI and Astral were to stay separate, independent companies. [3][4][14]
On the open source question, both sides made commitments. OpenAI said that, in keeping with a developer-first approach, it planned to support Astral's open source products after closing. Astral said OpenAI would keep supporting its open source tools and that the team would "keep building in the open, alongside our community," while exploring ways the tools could work more closely with Codex. Neither announcement said much about pyx, the commercial registry, which several observers flagged as a sign of uncertainty about that part of the business. [1][2][5]
The Astral deal was part of a sharp increase in dealmaking at OpenAI. After making no acquisitions before 2023, the company picked up the pace in 2025 and accelerated again in 2026. Crunchbase News reported that OpenAI had done roughly as many acquisitions in the first months of 2026 as in all of 2025, when it made eight. [4][15]
Several of the 2026 deals pointed at developer tooling specifically. The same month as Astral, OpenAI said it would buy Promptfoo, an open-source tool for testing and evaluating AI applications. Earlier deals in the run-up included the AI consultancy Convogo and the health records startup Torch, both in January 2026. The largest disclosed OpenAI acquisition remained its 2025 purchase of io, the AI hardware startup tied to designer Jony Ive, reported at about $6.5 billion. In 2025 the company also bought the product analytics firm Statsig for roughly $1.1 billion and made its founder a senior executive. In December 2025 OpenAI hired a corporate development leader away from Google, a signal that more deals were coming. [3][4][15]
The competitive backdrop mattered too. AI coding assistants had become one of the hottest segments in software, and OpenAI was racing for users against rivals such as Anthropic, whose Claude models are popular with developers, and Cursor, the AI code editor. Buying the team behind Python's most used tooling gave OpenAI both engineering talent and a foothold in infrastructure that sits underneath a large share of Python development. [3][4]
Reaction in the Python and open source world was mixed, with appreciation for Astral's work alongside worry about consolidation. A recurring concern was whether OpenAI bought Astral mainly for its tools or mainly for its people. The developer and writer Simon Willison noted that "a product+talent acquisition can turn into a talent-only acquisition later on," meaning the tools could be left to drift if the engineers were redirected to other work. [5]
A second concern was competitive leverage. Willison warned that "one bad version of this deal would be if OpenAI start using their ownership of uv as leverage in their competition with Anthropic," for example by tying the tooling more tightly to OpenAI's own products. The absence of clear commitments about pyx added to the unease, since the registry was the piece most likely to be steered toward commercial ends. [5]
Set against those worries was the protection offered by permissive licensing. uv is dual-licensed under the MIT and Apache 2.0 licenses, and Ruff is under the MIT license, so existing code cannot be clawed back. Several commentators argued that this made the worst case survivable. Willison cited Astral's Douglas Creager, who put it this way: "That makes the worst-case scenarios have the shape of 'fork and move on', and not 'software disappears forever'." He also pointed to the developer Armin Ronacher's view that uv is "very forkable and maintainable," so the community could carry it on if OpenAI's stewardship faltered. [5]
Tooling vendors took a measured line. In a post for its PyCharm editor, JetBrains said its "commitment to supporting the best Python tooling for our users stays the same," noting that it had been integrating Ruff and uv into PyCharm and would continue to do so regardless of who owned Astral. It acknowledged the risk that engineers could be pulled toward commercial priorities, while stressing that permissive licenses meant the tools could not go backwards, and that "the community can fork them if it ever comes to that." [16]
The deal mattered because of how central Astral's tools had become in a short time. In a few years, uv and Ruff went from new projects to default choices in many Python workflows, including continuous integration pipelines, container builds, and the setup steps of countless projects. Ownership of that layer by a single AI lab was a new situation for the ecosystem, which had historically leaned on community-run or foundation-backed tooling. [4][6]
For OpenAI, the purchase reflected a broader move from building models toward owning more of the software supply chain around them, so that Codex could not only write code but also manage dependencies, enforce quality, set up environments, and maintain projects over time. For Python users, the near-term practical effect was small, since the tools stayed free and open source, but the longer-term path now depended on choices made inside OpenAI. The permissive licenses left the community a fallback, and that fallback, the ability to fork, became the main reassurance people pointed to while the deal awaited closing. [1][5][16]