Cloudflare open-sources Forge, its homegrown SDK generator, after Anthropic shutters Stainless

Cloudflare has unveiled Forge, an open-source tool that transforms API definitions into ready-to-use SDKs, CLI tools, documentation, and other integration layers. Released under an Apache 2.0 license, the project runs on private infrastructure without requiring payment or reliance on Cloudflare's servers.

In a blog post, the company's engineering leads emphasize that SDK tooling is infrastructure too essential to depend on proprietary platforms: "We believe that building tools for APIs is a core part of the Internet, and you should be able to do that without needing a SaaS product."

The Stainless reckoning#

Cloudflare's move reflects broader industry frustration after Anthropic acquired Stainless in May and promptly shut down the service. Stainless had been a critical dependency for many API providers—including Cloudflare, Google, and OpenAI—to maintain their SDK libraries and update them as APIs changed.

Google signaled the same concern just 11 days before Cloudflare's announcement, partnering with Speakeasy to open-source an OpenAPI code-generation suite. Google's statement was direct: "This sudden disruption highlighted that proprietary, closed-source generators create unacceptable platform risk."

Why Cloudflare built its own#

Cloudflare's dissatisfaction with hosted generators predates the Stainless shutdown. Managing over 3,500 API operations across hundreds of services requires a generation system that stays synchronized without constant cross-team coordination. The company also wanted engineers to preview how API changes would cascade through SDKs, CLIs, and documentation before merging code—a capability external vendors struggled to provide.

Beyond traditional SDKs, Cloudflare needed generation to cover newer surfaces like MCP servers and Cap'n Web bindings, which allow JavaScript applications to call remote services like local functions. When issues emerged in one part of the API, they would surface only downstream, forcing engineers to debug through systems they didn't fully control and coordinate fixes across company and vendor boundaries.

The AI agent factor#

The stakes have risen with AI agents, which increasingly rely on SDKs and CLI bindings to understand a service's capabilities. Stale or incomplete interfaces don't just frustrate developers—they can mislead agents about what a service can do.