Vercel AI SDK is an open-source TypeScript toolkit for building AI-powered applications. It provides a unified API across providers like OpenAI, Anthropic, Google, and Mistral, with core functions for text and object generation, React/Next.js hooks, tool calling, and streaming support.¹⁾

The AI SDK Core provides primitive functions for LLM interactions:²⁾

• generateText – generates text from any supported model; suitable for Q&A, batch processing, classification
• streamText – streams text tokens in real-time for responsive chat UIs and progressive content generation
• generateObject – produces structured JSON output constrained to a Zod schema; eliminates brittle text parsing
• streamObject – streams structured objects incrementally for progressive UI updates with validated JSON

All functions use a unified API, requiring only provider/model swaps for portability.

Streaming is a first-class concern in the AI SDK:³⁾

• Token-by-token delivery for real-time user experiences
• Built-in handling of Server-Sent Events (SSE)
• Backpressure management and error recovery
• Edge runtime compatible for low-latency global deployment
• Works across Vercel Edge, Cloudflare Workers, and Node.js

The tool system enables AI models to execute functions and interact with external systems:⁴⁾

• Define tools with Zod-validated parameter schemas and execute functions
• Model decides which tool to call based on conversation context
• SDK validates inputs, executes the function, and passes results back to the model
• maxSteps enables automatic multi-step tool execution for agentic workflows
• Integrated with both generateText and streamText

The SDK provides a unified interface across all major LLM providers:⁵⁾

• OpenAI – GPT series models
• Anthropic – Claude models
• Google – Gemini models
• Mistral – Mistral and Mixtral models
• Cohere, Hugging Face, Amazon Bedrock – additional providers
• Switch providers by changing a single import line
• AI Gateway proxy adds retries, fallbacks, token tracking, and cost management

AI SDK UI provides React hooks for interactive applications:⁶⁾

• useChat – manages chat state, streaming messages, multimodal inputs, and tool calls
• useCompletion – streamlined hook for non-conversational text completion
• useObject – streaming structured data to React components
• Framework support for React, Svelte, Vue, and Angular
• Optimized for Next.js App Router and Server Actions

The SDK integrates directly with Next.js Server Actions:⁷⁾

• Server-side generation and streaming without separate API routes
• API keys remain secure on the server
• End-to-end type safety between server and client
• React Server Component streaming for AI-generated UI
• 70% reduction in API boilerplate compared to REST-based approaches

• v4 (2025) – introduced generateObject for structured outputs, multi-step tool calling with maxSteps, useObject for streaming structured data, and provider middleware for logging and caching⁸⁾
• v6 (2026) – shifted from REST routes to native Server Actions, unified API enhancements, multimodal support improvements, and agent optimizations⁹⁾
• The ai npm package reaches approximately 3.5 million weekly downloads as of March 2026

• Streamlit for AI
• Gradio
• Hugging Face