OpenAI acquires Astral behind uv and Ruff

OpenAI is buying Astral, the company behind widely used open-source Python tools including the uv package manager and the Ruff linter, and folding it into OpenAI’s Codex team. Ars Technica reports that uv alone now sees more than 120 million monthly downloads, while Ruff is downloaded roughly 180 million times a month—numbers that make Astral less a niche developer shop than a piece of the modern Python supply chain.

The deal’s financial terms were not disclosed. OpenAI said the acquisition will “accelerate our work on Codex” and let AI agents work more directly with tools developers already use. Astral founder Charlie Marsh said OpenAI will continue supporting the tools and “keep building in the open.”

The strategic logic is straightforward: controlling developer tooling is a distribution channel. Coding assistants compete on model quality, but they also compete on where they sit in the workflow—dependency management, formatting, linting, type checking, CI pipelines, and IDE defaults. Whoever owns the toolchain can decide what integrates first, what becomes the default button, and what breaks last.

Python is particularly sensitive to this because its productivity depends on managing environments and dependencies across projects. A fast package manager like uv is not just a convenience; it is the gate through which new libraries, security fixes and build systems flow. If an AI coding agent is tightly integrated into that gate, it can influence which runtimes, cloud services and APIs are “one command away,” and which require manual workarounds.

This is not unique to OpenAI. Ars notes that Anthropic acquired Bun, a JavaScript runtime, last year as it pushed Claude Code. The pattern is vertical integration: model providers buying the tooling layer that determines developer habits. The economics are attractive because tooling is sticky—teams standardise on it, build scripts around it, and train staff on it. Once embedded, switching costs rise, and pricing power can be exercised elsewhere: in hosted services, enterprise contracts, or proprietary integrations.

Open source complicates the story. The code remains visible, forks remain possible, and communities can resist. But ownership still matters because maintenance, release cadence, and roadmap decisions shape the “official” version most organisations will trust. The more central the tool, the more it becomes infrastructure—and infrastructure tends to pull toward the entity that can fund it, insure it, and bundle it.

For now, OpenAI is promising continuity. The test will be whether the most widely used Python tools remain neutral plumbing, or become the on-ramp to a particular AI stack.

Astral’s tools will keep shipping under open licences, but their default integrations will now be decided inside a model company.