Google and Speakeasy Open Source AI SDK Generation Tooling
Partnership addresses platform risks in proprietary client library generation
1Proprietary tooling risks
Google faced a critical challenge when their proprietary SDK generation provider was acquired and shut down during preparations for Google I/O 2026. This incident exposed the vulnerability of relying on closed-source tools for essential infrastructure. The abrupt disruption threatened the launch timeline for their Interactions API and highlighted how single-vendor dependencies can jeopardize platform stability. Such risks are particularly acute in fast-moving AI ecosystems where API specifications evolve rapidly.
2Migration under pressure
The engineering team had to rebuild their SDK pipeline on a tight schedule while minimizing developer disruption. They partnered with Speakeasy to migrate client libraries in place, requiring careful handling of type definitions across languages and preservation of error hierarchies. The migration maintained strict streaming behavior and integrated the new generator into Google's internal monorepo. This complex transition ensured continuity for existing API consumers while laying groundwork for future expansion.
3Technical requirements
Generating clean SDKs from OpenAPI specs demands determinism and strict type safety, especially for AI APIs handling complex protocols. The solution combines a fast, deterministic core generator with AI agents for custom parts. This hybrid approach maintains quality while reducing maintenance overhead from multiple engineers to approximately one. The system currently supports six targets including three released SDKs, with more in development pipeline.
4Open source commitment
Speakeasy is open sourcing its full OpenAPI client suite under AGPLv3 to prevent similar ecosystem disruptions. The license allows commercial use while keeping generator improvements open. Generated SDKs remain under users' chosen licenses (MIT/Apache 2.0). This model balances community benefits with commercial flexibility, addressing a critical gap in API tooling infrastructure for AI and cloud services.
5Engineering efficiency gains
The new system dramatically reduced maintenance overhead while expanding language support. Previously requiring multiple engineers for handcrafted generators, the current setup supports six targets with roughly one engineer. This efficiency comes from combining deterministic core generation with AI-assisted customization, demonstrating how strategic automation can scale critical developer infrastructure without proportional headcount growth.
6License strategy
The AGPLv3 license choice reflects careful consideration of commercial and community needs. While allowing free use in development/CI pipelines, it requires sharing improvements to the generator itself. This ensures the base technology remains open while letting organizations retain full ownership of generated SDKs. The approach mirrors successful open core models in other infrastructure software categories.
7Current implementations
Google has already deployed the new system for its Gemini Interactions API, demonstrating real-world viability. Developers can explore these updated SDKs today while the Speakeasy generator is available on GitHub for broader use. The solution supports generating client libraries, CLIs, and MCP servers from OpenAPI specs, covering diverse integration needs across modern software stacks.
8Ecosystem implications
This move establishes open source as the foundation for critical API infrastructure, reducing single-vendor risks across the industry. By open sourcing generation tooling while keeping SDKs proprietary, it creates a sustainable model for maintaining essential developer tools. The approach could influence how other major platforms handle similar infrastructure challenges in AI and cloud services.