Anthropic’s acquisition of Stainless, announced May 18, puts one AI lab in control of the SDK-generation pipeline that three of its largest competitors rely on to ship client libraries. OpenAI, Google, and Cloudflare now face a forced migration to alternative tooling, and the clock started the day the deal closed.
What Stainless does and who depended on it
Stainless, founded in 2022 by ex-Stripe engineer Alex Rattray, built an SDK generator that turned OpenAPI specs into idiomatic client libraries across TypeScript, Python, Go, Java, and other languages. The tool used generative AI for initial configuration and offered an SDK Studio for manual review, which separated it from older options like OpenAPI Generator. Rattray’s earlier work at Stripe, where he built the original codegen system using a JSX and Prettier pipeline, gave Stainless its technical DNA.
The customer list included OpenAI, Google, Cloudflare, Replicate, Runway, Together AI, Lithic, LangChain, and Orb. Cloudflare’s API Platform Tech Lead Jacob Bednarz said Stainless let the company auto-ship over 1,000 endpoints across three language SDKs. OpenAI’s Python SDK alone was pulling millions of downloads per week as of April 2024.
Every official Anthropic SDK was also built on Stainless from the start. The acquisition target was, in effect, Anthropic’s own infrastructure vendor that happened to serve its competitors.
The deal terms and what Anthropic said
The Information reported, per TechCrunch, that the acquisition price exceeds $300 million. Anthropic has not disclosed terms. Stainless was backed by Sequoia Capital and Andreessen Horowitz with a $150 million valuation as of December 2025, meaning the deal more than doubled that figure.
According to Analytics Insight, this is Anthropic’s fourth acquisition in six months, following Bun (the JavaScript runtime, December 2025), Vercept (computer operations), and Coefficient Bio (AI biotech).
The critical detail for anyone not named Anthropic: the company told TechCrunch it will wind down all hosted Stainless products, including the SDK generator. Existing customers retain ownership of SDKs they have already generated, with full rights to modify and extend them. But the tool that produced those SDKs is going dark.
Who loses access and what it costs them
OpenAI, Google, and Cloudflare carry the highest profiles among affected customers, but the long tail includes Replicate, Runway, Together AI, LangChain, Lithic, and Orb. None have made public statements about migration plans as of May 26.
The immediate consequence is SDK frost. Any company that relied on Stainless to keep its client libraries in sync with a changing API surface now has a generated-but-static artifact. New endpoints, changed parameter schemas, and deprecations will not propagate automatically. The work either moves to a different generator or becomes manual.
For OpenAI, the dependency is particularly visible. Millions of weekly Python SDK downloads means a significant installed base expecting that library to track the API surface. Google and Cloudflare have their own installed bases, though the volume is less publicly documented. Any of these companies can absorb the migration cost internally; the variable is timeline, and whether the replacement matches Stainless on output quality.
The alternatives: LibLab, Speakeasy, and build-in-house
The SDK-generator market is small enough that Stainless’s exit reshapes it. LibLab and Speakeasy are the closest commercial competitors, both offering SDK generation from API specs. The open-source OpenAPI Generator has been around longer and supports more languages, but historically required more manual tuning to produce idiomatic output.
The fourth option is build-in-house. Rattray did it at Stripe before Stainless existed. The engineering cost is not trivial; Cloudflare’s 1,000-plus endpoints across three languages gives a sense of the surface area. But for a company with OpenAI or Google’s resources, it is feasible. The tradeoff is time: every week spent building an internal codegen pipeline is a week the SDK falls further behind the API.
Why Anthropic wants Stainless
The stated rationale is MCP. Anthropic created the Model Context Protocol to standardize how AI agents connect to external tools and data sources. Katelyn Lesse, Anthropic’s Head of Platform Engineering, framed the acquisition in those terms: “agents are only as useful as what they can connect to.” Stainless’s SDK-generation capability feeds directly into Anthropic’s ability to produce MCP servers and client libraries for its own ecosystem faster than competitors can rebuild theirs.
The broader pattern is consistent. Anthropic has been buying developer-infrastructure companies at an accelerating pace: Bun gave it a JavaScript runtime, Vercept added computer-use capabilities, and now Stainless locks down the SDK-pipeline layer. Each acquisition removes a dependency on a third party and adds one for competitors.
What happens next
No competitor has announced a migration plan. The scope of the wind-down is ambiguous; it could mean the hosted Stainless Studio goes dark while the underlying generator continues under license, or it could mean the entire tool is absorbed into Anthropic’s internal stack with no external access whatsoever. The difference determines whether any team running Stainless under a self-hosted or enterprise agreement has a transition window or a hard cutoff.
What is clear is that the decision is now forced. Every lab that was on Stainless has to choose among LibLab, Speakeasy, OpenAPI Generator, or an internal build. The longer that decision takes, the more their SDKs drift from their APIs. The acquisition did not remove Stainless from the market as much as it started a countdown for everyone who was not Anthropic.
Frequently Asked Questions
Does the Stainless wind-down affect only SDKs, or also CLIs and MCP servers?
Stainless powered generation of CLIs and MCP servers in addition to REST client libraries. The wind-down of all hosted products means external teams lose the ability to generate any of these artifacts through Stainless, not just SDKs. For Anthropic’s MCP roadmap, the CLI and MCP-server generation capabilities are likely being absorbed internally, but no public statement confirms which specific Stainless outputs will survive in any form.
Why can’t affected companies simply switch to OpenAPI Generator and move on?
OpenAPI Generator is fully open source, so no vendor can cut off access the way Anthropic just did. The tradeoff is output quality: it lacks any equivalent to Stainless’s AI-assisted config generation or its SDK Studio review interface, and it produces less idiomatic client code by default. Teams migrating from Stainless will likely need to write custom Handlebars templates or post-processing scripts to close that quality gap, adding weeks of engineering work that Stainless handled automatically.
How does a $300M deal compare to Anthropic’s overall spending capacity?
Anthropic was valued at $380 billion as of February 2026, making the Stainless acquisition less than 0.1% of that figure. The four-deals-in-six-months pace (Bun for JavaScript runtime, Vercept for computer-use, Coefficient Bio for biotech, Stainless for SDK generation) shows a systematic layer-buying pattern across adjacent dev-infrastructure categories. Each deal is small enough individually to avoid mandatory antitrust review, but the cumulative pattern of acquiring tools that competitors depend on could trigger a pattern-based FTC or DOJ examination.
What specific compounding risk does SDK frost create for high-download libraries?
When an SDK stops auto-syncing with its API surface, every new endpoint, renamed parameter, or deprecated field becomes a silent divergence between what the SDK exposes and what the API actually accepts. Downstream applications start hitting undocumented behaviors or missing features, and the support burden shifts from the codegen tool to manual documentation updates and issue-triage. The longer a company like OpenAI operates without a replacement pipeline, the wider that gap grows and the harder the eventual catch-up migration becomes, because every accumulated divergence must be reconciled at once.