TanStack, the team behind popular TypeScript libraries like TanStack Query and TanStack Table, has released the alpha version of TanStack AI, a framework-agnostic AI toolkit designed to eliminate vendor lock-in and provide developers with control over their AI stack.
TanStack AI introduces a unified interface across multiple AI providers, multi-language server support, and an open protocol architecture. The alpha ships with JavaScript/TypeScript, React, and Solid support, along with adapters for OpenAI, Anthropic, Gemini, and Ollama. The release represents a different approach to AI tooling, positioning itself as vendor-neutral infrastructure rather than a platform service.
One of the standout features in TanStack AI is its isomorphic tool system, which allows developers to define tools once using toolDefinition() and provide environment-specific implementations with .server() or .client() methods. This architecture delivers type safety across the entire application while enabling tools to execute in both server and client contexts.
There are two different approaches to define the tool schemas, either via Zod Schemas, which is the reccomended approach, or using JSON Schema, which may be useful for projects that already have JSON Schema Definitions. The toolkit provides per-model type safety that gives developers full typing for provider-specific options on a per-model basis.
Client libraries include vanilla JavaScript, React, and Solid, with additional frameworks planned. The alpha also ships with isomorphic devtools that provide insight into LLM behavior on both server and client sides, allowing developers to debug AI workflows using familiar patterns.
The release has generated positive reactions from the developer community. Developer Stanley Ulili wrote in a detailed Better Stack guide:
It is still in alpha, but it already looks promising. The focus is on clean architecture, strong TypeScript support, and the freedom to fit into your current stack instead of forcing a specific framework or provider
Over on reddit, a number of commenters have raised questions around the use cases for the SDK and what this new library is aiming to solve, which prompted a response from Tanner Linsley, the founder of the TanStack ecosystem:
I and all of the other maintainers at TanStack are really having a lot of fun with AI lately and we saw enough improvement space on Vercel’s solution that we wanted to build our own. One that is as close to our [product tenets](https://tanstack.com/tenets) as possible.
So far, that’s resulted in better type-safety, better patterns around isomorphism, and honestly just the freedom to move in the direction we want without being beholden to another team.
Competition is good. Breeds greatness all-around.
TanStack AI positions itself as a direct alternative to the Vercel AI SDK, the dominant player in the JavaScript AI toolkit space. Unlike Vercel’s approach, TanStack AI operates as pure open source infrastructure with no service layer, no platform fees, and no vendor lock-in. The team emphasizes that developers connect directly to their chosen AI providers rather than through a middleman.
Since this is an alpha release of a new library, there is no migration path from earlier versions. Developers can get started by installing the core packages via npm: npm install @tanstack/ai @tanstack/ai-react @tanstack/ai-openai. The quick start guide provides step-by-step instructions for creating a chat application, while the tools guide explains the isomorphic tool system in depth.
TanStack AI is an open source project developed and maintained by the TanStack team. It builds on the team’s established reputation for creating framework-agnostic developer tools, with the goal of providing honest open source tooling that works with any stack rather than forcing developers into proprietary ecosystems.
