Vercel AI SDK

January 30, 2026

The Vercel AI SDK (also known as the AI SDK) is a TypeScript library that provides a unified API for working with AI models from multiple providers. Write once, swap providers freely.

Core Proposition

Instead of learning provider-specific APIs:

// Before: Provider-specific
import OpenAI from 'openai';
import Anthropic from '@anthropic-ai/sdk';
// Different APIs, different patterns...

Use one consistent interface:

import { generateText } from "ai";

const { text } = await generateText({
  model: "anthropic/claude-sonnet-4.5",
  prompt: "What is love?",
});

The SDK handles provider differences behind a consistent abstraction.

Key Features

Provider Agnostic

Swap between providers by changing a string:

• OpenAI (openai/gpt-4o)
• Anthropic (anthropic/claude-sonnet-4.5)
• Google (google/gemini-2.0-flash)
• Mistral, Cohere, and more

Streaming First

Built for real-time AI interactions:

import { streamText } from "ai";

const stream = await streamText({
  model: "anthropic/claude-sonnet-4.5",
  prompt: "Write a poem about coding",
});

for await (const chunk of stream) {
  process.stdout.write(chunk);
}

Tool Calling

Function calling with type safety:

import { generateText, tool } from "ai";
import { z } from "zod";

const result = await generateText({
  model: "openai/gpt-4o",
  tools: {
    weather: tool({
      description: "Get weather for a location",
      parameters: z.object({
        city: z.string(),
      }),
      execute: async ({ city }) => {
        return { temp: 72, condition: "sunny" };
      },
    }),
  },
  prompt: "What's the weather in Tokyo?",
});

Structured Output

Parse model responses into typed objects:

import { generateObject } from "ai";
import { z } from "zod";

const { object } = await generateObject({
  model: "openai/gpt-4o",
  schema: z.object({
    name: z.string(),
    age: z.number(),
    hobbies: z.array(z.string()),
  }),
  prompt: "Generate a fictional person profile",
});

Agent Support (v5+)

Version 5 introduced experimental agent primitives:

import { experimental_createAgent as createAgent } from "ai";

const agent = createAgent({
  model: "anthropic/claude-sonnet-4.5",
  tools: { /* ... */ },
  system: "You are a helpful assistant",
});

const result = await agent.run("Research the latest AI news");

As noted on Twitter:

ai-sdk-tools/agents orchestrates multiple specialized agents with automatic routing, seamless handoffs, and preserved context.

Multi-Agent Orchestration

Community extensions enable sophisticated patterns:

// Specialized agents for different tasks
const researchAgent = createAgent({ /* research-focused */ });
const analysisAgent = createAgent({ /* analysis-focused */ });
const writingAgent = createAgent({ /* writing-focused */ });

// Orchestrate with handoffs
const result = await orchestrate({
  agents: [researchAgent, analysisAgent, writingAgent],
  task: "Write a market analysis report",
});

Workflow Patterns

The SDK supports common orchestration patterns:

• Sequential: Chain model calls
• Parallel: Run independent calls simultaneously
• Conditional: Branch based on outputs
• Agentic loops: Iterate until completion

React Hooks

For frontend integration:

import { useChat } from "ai/react";

function ChatUI() {
  const { messages, input, handleInputChange, handleSubmit } = useChat();
  
  return (
    <div>
      {messages.map(m => <div key={m.id}>{m.content}</div>)}
      <form onSubmit={handleSubmit}>
        <input value={input} onChange={handleInputChange} />
      </form>
    </div>
  );
}

When to Use It

Good fit:

• TypeScript/JavaScript projects
• Need provider flexibility
• Building Next.js or React apps
• Want streaming out of the box
• Need structured output parsing

Consider alternatives when:

• Python-only codebase
• Need low-level provider control
• Complex multi-agent systems (consider LangGraph)

Installation

npm install ai @ai-sdk/openai @ai-sdk/anthropic

import { generateText } from "ai";
import { openai } from "@ai-sdk/openai";
import { anthropic } from "@ai-sdk/anthropic";

Sources

• AI SDK Documentation — Official docs
• Vercel Blog — AI SDK announcements
• GitHub: vercel/ai — Source code

See also: TanStack AI · LangChain · Orchestration