Anthropic Acquires Stainless for $300M+: The SDK Factory Behind OpenAI and Google

Anthropic has acquired Stainless, the SDK generation startup that writes the official developer libraries for OpenAI and Google, for more than $300 million — and it is immediately shutting down the hosted product for every customer except itself.

That last sentence carries the strategic weight. Stainless wasn’t a general-purpose tool that Anthropic wanted to own. It was an infrastructure layer that OpenAI, Google, and Cloudflare all depended on to maintain production-quality SDKs across Python, TypeScript, Go, Kotlin, and Java. By acquiring it and winding down external access, Anthropic has removed a piece of critical developer infrastructure from its competitors’ hands.

The deal closed on May 16, 2026. The implications for every developer building on any major AI API are significant and worth understanding in detail.

What Stainless Actually Does — and Why It’s Hard to Replace

Stainless was founded in 2022 by Alex Rattray, who built key aspects of Stripe’s API developer experience, including the patented codegen system that powers Stripe’s API client libraries. Backed by Sequoia Capital and Andreessen Horowitz, the New York startup built an AI-powered compiler that takes an OpenAPI specification and automatically generates production-ready software development kits across multiple languages.

That description undersells what the product actually does. Anyone who has built and maintained SDKs across five programming languages understands the operational burden: every SDK needs retries with exponential backoff, streaming response handling, pagination, rate limit management, error hierarchies, type safety, and exhaustive test coverage. Every API change — new endpoint, new parameter, deprecated field — needs to be propagated consistently across all language variants without breaking existing integrations.

Doing this manually for one language is tedious. Doing it for five languages simultaneously, with production users depending on the libraries, is a significant engineering resource commitment. Stainless automated it. An API spec change propagates to all SDKs automatically, complete with the retries, streaming, pagination, and error handling that make a library actually production-grade rather than a thin API wrapper.

# What Stainless generates from a single OpenAPI spec

openai/openai-python     → Python SDK     (pip install openai)
openai/openai-node       → TypeScript SDK (npm install openai)
openai/openai-go         → Go SDK
openai/openai-java       → Java SDK
openai/openai-kotlin     → Kotlin SDK

All with retries, streaming, pagination, type safety, and tests baked in.
Each generated from the same spec — updated automatically on every API change.

That’s the OpenAI SDK repository list. Every one of those libraries was generated and maintained using Stainless. Google’s Gemini API SDKs used the same tooling. Cloudflare’s API libraries. Anthropic’s own SDKs, according to Anthropic, “since the earliest days of our API.”

The Strategic Logic: Developer Infrastructure as Competitive Moat

The acquisition price of more than $300 million for a four-year-old startup with roughly 20 employees points to strategic value well above what the company’s current revenue would justify on a pure financial basis. What Anthropic is buying is not a SaaS product — it’s infrastructure control at a critical layer of the developer toolchain.

SDK quality is a significant factor in API adoption. Developers evaluate APIs partly on documentation, partly on pricing, and significantly on the quality of the SDK for their preferred language. A production-grade Python SDK with proper async support, type hints, and streaming means a developer can go from API key to working code in minutes. A poorly maintained SDK with missing features or inconsistent error handling can be a genuine barrier to adoption even when the underlying API is excellent.

Stainless’s tooling gave both OpenAI and Google a consistent, high-quality SDK layer without allocating dedicated engineering teams to maintain it. Those teams must now build or acquire that capability independently. And they must do it from scratch — Anthropic will use Stainless internally, and the hosted platform is shutting down. There is no migration path or extended transition period being offered to competitors.

This is an unusual move. Most enterprise software acquisitions include a wind-down period for existing customers. Stainless is cutting access immediately. The message to OpenAI and Google is unambiguous: the infrastructure you relied on is now owned by your direct competitor and is no longer available to you.

The MCP Angle: Stainless Was Building More Than SDKs

The timing of the acquisition relative to the Model Context Protocol’s explosive growth is not coincidental. Stainless had been expanding its tooling to cover MCP server generation alongside traditional API SDKs. With MCP now the de facto standard for connecting AI agents to external tools — 97 million downloads and integration into every major AI development platform — the company that automates MCP connector generation has a significant position in the agentic AI infrastructure stack.

Anthropic created MCP and has the most direct stake in its adoption. Controlling the tooling that makes it easiest to generate high-quality MCP servers accelerates the ecosystem around Claude-based agent development. Every developer who builds an MCP connector using Anthropic’s Stainless-powered tooling is building deeper into the Claude ecosystem. This is the same flywheel logic that made the original Stainless investment attractive to Sequoia and Andreessen Horowitz — SDK infrastructure compounds over time as API usage and ecosystem complexity grow.

What Happens to Existing Stainless Customers

The practical situation for developers who were using Stainless directly is nuanced. Anthropic has confirmed that customers retain full ownership of all SDKs they’ve generated to date. Existing generated code is not going away — you own your output. What is going away is the ability to continue using Stainless’s hosted platform to generate new SDKs or automatically maintain existing ones as your API evolves.

For OpenAI, Google, and Cloudflare, the most immediate problem is SDK maintenance cadence. Every API update that previously triggered automatic multi-language SDK updates now requires manual intervention or a new toolchain. The open-source SDKs already exist; the automation that kept them current does not.

For smaller companies and independent developers who were using Stainless to generate SDKs for their own APIs, the available alternatives are:

• Fern: the most direct Stainless competitor with AI-assisted SDK generation across TypeScript, Python, Go, Java, and Ruby. Likely to absorb significant inbound demand following the acquisition announcement.
• Speakeasy: another SDK generation platform with TypeScript, Python, Go, and Java support and an active enterprise customer base.
• OpenAPI Generator: open-source, broad language support, but generates thinner code without Stainless’s production-grade features like retries and streaming built in automatically.
• Manual build: appropriate for simple APIs or teams with dedicated SDK engineering resources willing to own the maintenance burden.

The competitive landscape for SDK generation will consolidate and accelerate following this acquisition. Fern and Speakeasy are the most obvious beneficiaries — neither has Stainless’s production history with frontier AI API providers, but both have working products that existing Stainless customers can migrate to immediately.

What OpenAI and Google Do Next

Both OpenAI and Google are large enough to build or acquire replacement SDK tooling. OpenAI has the engineering resources to build internal SDK generation infrastructure, and the MCP ecosystem means there are now open-source tools that partially automate the connector generation problem. Google, with its extensive internal developer tooling history, has comparable options.

The more acute problem for both is timeline. Stainless allowed them to move fast on SDK maintenance without allocating dedicated headcount. Building a replacement takes months. During that period, SDK updates will be slower and more manual. For most developers, this will be invisible — existing SDKs continue to work. For developers tracking new API features closely, the latency between API update and SDK support may increase.

There is also the question of whether either company tries to counter-acquire in the SDK generation space. Fern has been building for several years and has enterprise customers. A defensive acquisition by OpenAI or Google to lock in their own SDK infrastructure would be a direct response to Anthropic’s move. Watch for acquisition activity in developer tooling over the next six to twelve months.

The Bigger Picture: Infrastructure Acquisitions as AI Strategy

The Stainless acquisition fits a pattern that has been emerging in AI infrastructure competition. As frontier model capabilities converge — all major labs are now operating at similar performance tiers for most common tasks — the differentiation increasingly comes from the developer experience layer. Which API is easiest to integrate? Which SDK is most production-grade? Which tooling makes it fastest to go from experiment to production deployment?

Anthropic’s $50 billion funding round earlier this year positioned it to make exactly these kinds of strategic infrastructure moves. The Stainless acquisition at $300M+ is small relative to Anthropic’s capitalization but disproportionately significant in developer ecosystem terms. SDK tooling is unsexy infrastructure that most developers never consciously think about — which is precisely why controlling it creates durable advantage.

The parallel to OpenAI’s DeployCo launch is instructive. DeployCo positions OpenAI at the enterprise delivery layer — embedding engineers inside organizations to build production AI systems. Stainless positions Anthropic at the developer infrastructure layer — controlling the tooling that makes it easiest to build production-grade integrations with Claude. Both moves reflect the same underlying logic: frontier model capabilities are converging, and the next phase of competition is won by owning the adjacent layers that developers and enterprises depend on daily.

Practical Guidance for Developers Building on AI APIs

If you’re building on any of the affected AI APIs, here’s what actually matters for your immediate workflow:

If you’re building on Anthropic’s Claude API: the existing Claude SDKs continue to work and will be maintained and improved by the combined Anthropic and Stainless team. The tooling that generated them is now internal, which should mean faster iteration on SDK quality, not slower. The Claude Managed Agents developer guide covers current Claude SDK patterns in detail.

If you’re building on OpenAI’s API: existing SDKs are stable and not going away. New feature support may come to SDKs slightly slower during the period before OpenAI builds replacement tooling. For most production applications, this has no immediate practical impact.

If you’re building your own API and were considering Stainless for SDK generation: evaluate Fern or Speakeasy as alternatives. Both are actively developed and will be absorbing significant inbound interest from former Stainless customers over the coming months. Act now rather than waiting for the hosted Stainless platform to degrade.

If you’re thinking about MCP server development: watch for Anthropic announcements about how Stainless’s MCP tooling will be surfaced through official Claude developer resources. The combined team is likely to prioritize MCP tooling given its strategic importance to the Claude agent ecosystem.

What This Tells You About Where AI Competition Is Headed

The Stainless acquisition signals that the next phase of AI competition will be fought at the infrastructure layer, not the model layer. When every major lab can train a model that scores at the frontier on standard benchmarks, the question of which model to use becomes secondary to the question of which ecosystem is easiest to build in, most reliable to operate in, and most cost-effective to scale.

For developers, this is net positive in the medium term. More competition at the developer experience layer means better tooling, better documentation, better SDK quality, and more investment in the resources that make it easier to go from working prototype to production deployment. The short-term disruption of Stainless’s hosted product shutting down is real but bounded. The longer-term outcome of intensified developer tooling investment from all major AI labs benefits everyone building on these platforms.

The $300 million that Anthropic paid for Stainless is a bet that SDK infrastructure is worth owning. Given that every official Anthropic SDK has been generated using Stainless since the beginning, and given that the MCP ecosystem is now the default connectivity layer for agentic AI, that bet looks well-constructed. The question is whether OpenAI and Google move to acquire equivalent infrastructure depth before Anthropic compounds the advantage further.

Developer tools, starter kits, and production templates for building on Claude’s API are available at wowhow.cloud.