Vercel Connect now supports TanStack AI

Vercel Connect has introduced support for TanStack AI, allowing agents built with the framework to interact with OAuth-protected MCP servers without requiring developers to manage or rotate credentials. The new integration introduces the @vercel/connect/tanstack-ai subpath, which exports a function called connectMCPTransport. This function accepts a TanStack transport configuration and attaches a Connect-backed authentication provider, ensuring tokens remain fresh by being refreshed before each MCP request.

The integration simplifies the process of connecting MCP clients to services like Linear, where per-user tokens are required. When a user has not granted access, the system automatically handles consent challenges by redirecting to Connect's consent URL. Developers can catch these challenges using getConsentChallenge and manage the flow programmatically, preventing consent errors from reaching the model as error strings.

To implement this, developers can follow the TanStack AI integration documentation provided by Vercel, which outlines the necessary steps to configure and deploy the integration. The documentation also includes guidance on supported frameworks and adapters, ensuring compatibility with existing Vercel Connect setups.

The new capability is designed to reduce the complexity of managing OAuth flows in agentic applications, allowing developers to focus on building functionality rather than handling authentication infrastructure. By offloading token refresh and consent management to Vercel Connect, the integration aims to improve both security and developer experience.