If you’ve used the OpenAI Python library, or the Google Gemini TypeScript SDK, or the Cloudflare AI client — you’ve used software generated by Stainless. The four-year-old startup founded by former Stripe engineer Alex Rattray built a system that takes an OpenAPI specification and spits out idiomatic, maintained, well-documented SDKs in TypeScript, Python, Go, Java, Kotlin, and more. Anthropic itself has used Stainless to generate every official Claude API library since the earliest days.
Anthropic announced on May 18 that it is acquiring Stainless for a reported $300M+, and that the hosted SDK generator will be wound down. Existing customers keep the SDKs they’ve already generated with full rights to modify and extend them — but the automated generation and update service goes away.
The practical consequence is that OpenAI, Google, Replicate, Runway, and the many other companies that relied on Stainless to maintain multi-language client libraries without a dedicated team of SDK engineers now have to figure out how to do that themselves. Writing an SDK once is fine; keeping it in sync with an evolving API across six languages without missing edge cases is the unglamorous engineering work that Stainless was handling. That work doesn’t disappear — it just moves back onto competitors’ plates.
The strategic logic here isn’t subtle. AI labs are competing for developer mindshare, and the SDK is the first thing a developer touches when they try an API. A well-maintained library with intuitive error handling, strong typing, and up-to-date method signatures creates a low-friction onboarding experience. A stale or inconsistent one is a quiet deterrent. Stainless was the shared infrastructure making many of those experiences good. Now it won’t be shared.
The stated reason for the acquisition leans on Model Context Protocol. Anthropic’s head of platform engineering: “Agents are only as useful as what they can connect to.” Stainless generates not just SDKs but also CLIs and MCP servers from API specs. As the ecosystem shifts toward agents that call external tools through MCP, the ability to generate correct, up-to-date connectors at scale becomes infrastructure-level important. Anthropic wants to own that layer.
There’s a reasonable counter-read here: Stainless’s technology isn’t magic. Generating SDKs from OpenAPI specs is a solvable engineering problem, and several alternatives exist (Speakeasy, liblab, Fern, and the older swagger-codegen lineage). The companies losing access to Stainless aren’t necessarily stranded — they have options and existing generated code to fork. But they lose the team, the institutional knowledge, and the ongoing refinement that Stainless had applied specifically to AI API patterns. The quality gap will be felt gradually, not all at once.
What Anthropic gains more concretely is Rattray and the Stainless engineers embedded in their platform team, working on MCP server generation and Claude’s connectivity surface. That’s the real asset — the team that thought deeply about what makes a developer experience good at the SDK layer, now focused entirely on making Claude easier to connect to the world.
Rattray’s framing — “SDKs deserve as much care as the APIs they wrap” — is the underlying thesis. In the agent era, where Claude is routinely calling dozens of external services in a single session, the quality of those connectors determines how reliably agents actually complete tasks. Anthropic is betting that owning this layer matters enough to pay $300M to foreclose the competition from the same tooling.
It’s an infrastructural bet as much as a competitive one, and a signal about where Anthropic thinks the platform leverage in the agent ecosystem actually sits.