SDK generation is a solved problem in theory. In practice, most generated SDKs are brittle wrappers that fall apart under real usage.

What it is: Stainless generates idiomatic, production-grade SDKs and MCP servers from an OpenAPI spec, and keeps them synchronized as the API evolves.

The core tension here is quality vs. automation. Hand-written SDKs are idiomatic and robust. Generated ones are fast to produce but tend to break language conventions, handle errors inconsistently, and fall out of sync the moment the underlying API changes. Stainless takes the position that generation and quality aren't a tradeoff. The output is meant to read like a senior engineer wrote it for each target language.

What makes it different: The MCP story is what caught my attention. Most MCP server implementations expose one tool per API endpoint. That approach doesn't scale: a large API floods the agent's context window with hundreds of schema definitions. Stainless takes a code-mode approach instead. The MCP server exposes two tools: search_docs and execute. Agents search documentation, then write and run TypeScript code against the generated SDK in a sandboxed Cloudflare Worker. The result is a reported 94-97% task accuracy, 3x fewer tool calls, and 100k+ tokens saved on complex tasks.

Key features:

SDK generation across TypeScript, Python, Go, Java, Kotlin, Ruby, C#, PHP, Terraform and more
Code-mode MCP server generated from the same OpenAPI spec
Auto-pagination, typed errors, and idiomatic parameter handling baked in
Automated GitHub workflows regenerate SDKs and MCP servers when the spec changes
Docs platform that stays in sync with the generated SDK
Secure sandbox execution via isolated Cloudflare Workers per request

Benefits:

API teams ship SDK updates without manual maintenance cycles
AI coding agents integrate with your API using SDK patterns rather than raw HTTP guesswork
Token consumption stays bounded regardless of API surface size
Fewer integration bugs at the point where agents write code against your API

Who it's for: Developer platform engineers and API product teams at companies that ship public APIs and need to maintain SDKs across multiple languages without a dedicated SDK team for each one.

The Anthropic and OpenAI SDKs were both built with Stainless, which tells you something about the bar it's hitting. As AI agents become a primary API consumer alongside human developers, the interface layer between API specs and agents is going to matter a lot more. Stainless is sitting at that intersection.

I hunt the latest and greatest launches in tech, SaaS and AI, follow to be notified.