ai

It provides a unified API for streaming text generation, tool calling, and multi-step agent workflows across providers including OpenAI, Anthropic, Google, Mistral, and dozens of others abstracting away provider-specific differences so application code doesn't need to change when switching models.

The SDK's streaming primitives are built for modern web frameworks, with first-class support for React Server Components and the App Router in Next.js, as well as compatibility with Node.js, Edge runtimes, and frameworks like SvelteKit, Nuxt, and Remix.

The useChat and useCompletion React hooks handle streaming state management on the client side, making it straightforward to build real-time chat interfaces, streaming text displays, and progressive UI updates without manually managing WebSocket connections or SSE streams.

For agentic applications, the AI SDK provides tool calling with automatic schema validation (via Zod), multi-step agent loops, and structured output generation the building blocks for AI assistants that can browse the web, call APIs, or manipulate data as part of a reasoning loop.

The SDK includes a Model Context Protocol (MCP) client for connecting to MCP servers, positioning it as a foundation for the emerging ecosystem of AI tool integrations. It is free and open-source under the Apache 2.0 license, with active development and comprehensive documentation maintained by Vercel.