Updated

AI SDK

Provider-agnostic TypeScript toolkit for streaming, tool calls and chat UIs.

5 makers' products we track use AI SDK. For AI SDK & agent framework it's 4th most chosen of 7 options: 5 of the 115 makers' products we know the AI SDK & agent framework of (4%). 241 open-source projects also declare it in their code. What makers like most: One TypeScript API across many model providers, so swapping models needs no code rewrite. Their most common complaint: Easy to drop in but hard to make resilient: no resumable streaming after a dropped connection and no multi-device support. On GitHub, 4 public pull requests moved projects to AI SDK from tools we list, and 3 moved them away.

Open sourceSelf-hostableFree tier

Profile checked

Ask your AI about this, with this page as the source:ChatGPT ↗Claude ↗Perplexity ↗

How makers use AI SDK

Makers' products we track use AI SDK for AI SDK & agent framework (5).

What AI SDK is used for

AI SDK & Agent Framework · Thin SDK →TypeScript agents with tool calling, multi-step loops and streaming to the UI, on any model provider.Trade-off: Memory, workflows and evals are left to you or other libraries.
LangChain66Mastra24LlamaIndex16AI SDK5CrewAI3Makers' products on each, for AI SDK & agent framework

AI SDK vs the alternatives makers chose

For AI SDK & agent framework, 5 of the 115 makers' products we know the choice of picked AI SDK (4%), behind LangChain (57%).

OptionMakers' productsOpen source
LangChain66 57%Yes
Mastra24 21%YesCompare →
LlamaIndex16 14%Yes
AI SDK5 4%Yes
CrewAI3 3%Yes
LangGraph1 1%Yes

For AI SDK & agent framework: how many of the 115 makers' products we know the AI SDK & agent framework of chose each option.

When to use AI SDK, and when not

Pick AI SDK if

Something else fits better if

  • You want a small typed agent loop in Python, close to the raw API: Pydantic AI or OpenAI Agents SDK.
  • Your agent works on files and runs commands, like a coding or research agent: Claude Agent SDK.
  • You need many prebuilt connectors for models, tools and data sources: LangChain.
  • Your agent is a long-running workflow with branches, saved state and human approval steps: LangGraph.
  • The agent mostly retrieves and reasons over your own documents: LlamaIndex.
From each decision's guide, written from what makers' products chose and why.

The AI SDK is Vercel's open-source TypeScript library for calling LLMs and building chat interfaces and agents on top of them. It puts one interface over the different provider APIs, so moving from one model vendor to another is a change of model, not a rewrite.

More about AI SDK

It has three parts. AI SDK Core has functions such as generateText and streamText, schema-checked structured output, tool calling, and a ToolLoopAgent class that runs the model-and-tools loop with stopping conditions. Providers are separate packages (@ai-sdk/openai, @ai-sdk/anthropic and many more) that implement a shared specification; a plain model string like anthropic/<model> goes through Vercel AI Gateway unless you set a different default provider. AI SDK UI has hooks (useChat, useCompletion, useObject) for React, Vue, Svelte, Angular and SolidJS. A third surface runs existing agent harnesses such as Claude Code and Codex. Core runs in Node.js and Deno as well as inside Next.js, Nuxt and SvelteKit.

The UI hooks handle streamed messages and loading and error state, and they read a documented stream protocol, so the backend can be written in another language, such as Python with FastAPI. Agents can connect to MCP servers over HTTP. It is a library with no hosted service required: it deploys wherever your JavaScript runs, and the Gateway is optional.

It is a toolkit, not a full agent platform: storing chat history is something you implement around useChat and streamText. The migration guides list four major versions, 4.0 through 7.0, and APIs prefixed experimental_ can change in any release.

How AI SDK itself is built

2 tools, from its own code, website and Product Hunt page.

AI SDK pros and cons

What makers like most: One TypeScript API across many model providers, so swapping models needs no code rewrite. Their most common complaint: Easy to drop in but hard to make resilient: no resumable streaming after a dropped connection and no multi-device support.

Pros

  • One TypeScript API across many model providers, so swapping models needs no code rewrite. PHPH 2HN
  • Streaming and tool calling are dependable, with tight Next.js and React integration. PHHNPH 2
  • A well-written core that other agent libraries build on, with version-matched docs shipped in the package. HNHN 2HN 3HN 4

Cons and watch-outs

  • Easy to drop in but hard to make resilient: no resumable streaming after a dropped connection and no multi-device support. HNHN 2
  • Streaming tool-call JSON is re-parsed on every chunk, which stalls on large payloads. HN

Themes that recur in makers' words and Hacker News comments, each linked to what it summarises, with how Product Hunt tags its reviews.

5.0★ from 5 reviews on Product Hunt ↗

Reliability and open issues

The most wanted open issue on GitHub has 66 👍.

Most wanted on GitHub

Open on 2026-10-04 in vercel/ai, with activity in the last year — issues and feature requests by 👍.

Who switches

More projects move to AI SDK: 4 pull requests in, 3 out; the most come from LangChain and leave for LangChain.

Moving to AI SDKMoving away
AI SDK

Public pull requests on GitHub since Oct 2024 whose title says "X to Y" — real code changes moving a project from one tool to another, by developers in general.

Who uses it

5 makers' products, each linked to the source that shows it, and 241 open-source projects that declare it in their code. The best known on Product Hunt: onlook (1,089 followers), Quartr (573 followers), Fuser (487 followers), Sparks AI (204 followers), Zyntax IDE (55 followers).

The maker says so 4Job posting 1Declared in code 242How evidence is collected →
onlook
onlookOpen-source visual editor for React apps: designers edit the UI and the changes are writ…

“• TypeScript-first, ergonomic API with excellent Next.js integration; streaming and tool calling are rock solid. • Consistent abstractions across multiple model providers (great with OpenRouter) minimize vendor lock-in and churn. • Clear docs and examples reduce boilerplate and accelerate iteration; fewer edge-case failures in production.”

AI SDK & Agent FrameworkMaker says so +1 · source ↗
Quartr
QuartrGo beyond numbers

“Used for multi-agent orchestration, state management, debugging, and structured artifact streaming for our AI application.”

AI SDK & Agent FrameworkMaker says so · source ↗
Sparks AI
Sparks AIA platform for your agents, teammates & everything you build

“AI SDK made super simple to integrate all the different models into our app. Its the best sdk for build ai apps right now.”

AI SDK & Agent FrameworkMaker says so · source ↗
Zyntax IDE
Zyntax IDECode Editor, Terminal, Git, AI Agent for Android

“I liked it because its a unified TypeScript SDK for building AI apps. Helped building Zyntax Agent.”

AI SDK & Agent FrameworkMaker says so · source ↗
aci
aciACI.dev is the open source tool-calling platform that hooks up 600+ tools into any agent…AI SDK & Agent FrameworkIn its code · source ↗
dory
doryAI-native SQL client for humans and agents. Query, explore, visualize, and open agent da…AI SDK & Agent FrameworkIn its code · source ↗
Dub
DubOpen-source link management and attribution platform for short links, conversion trackin…AI SDK & Agent FrameworkIn its code · source ↗
Dyad
DyadFree, open-source desktop AI app builder for macOS, Windows and Linux that runs locally…AI SDK & Agent FrameworkIn its code · source ↗
elmo
elmoOpen-source Answer Engine Optimization (AEO) and Generative Engine Optimization (GEO) pl…AI SDK & Agent FrameworkIn its code · source ↗

Open source: a project that declares AI SDK as a dependency in its public code — verifiable, but not necessarily a live product.

What its projects tell coding agents

122 of the 155 open-source products built with AI SDK that were worked on in the last 180 days keep instruction files for coding agents — 79%, against 53% of all the projects we read. AGENTS.md is in 102, CLAUDE.md in 74.

1,135words in the median file
pnpmthe package manager most named, in 40 of 112 files
60%say how to test a change
Named more in these files than in other projects' — in how many of the 112:
What AGENTS.md files say about other tools →What open-source products tell their coding agents →

Alternatives to AI SDK

All alternatives by situation →

Further reading

Questions makers ask about AI SDK

Is AI SDK good for production?

5 makers' products we track run on it. On GitHub, 4 public pull requests moved projects to AI SDK from tools we list, and 3 moved them away. The watch-out makers raise most: Easy to drop in but hard to make resilient: no resumable streaming after a dropped connection and no multi-device support.

What is AI SDK used for?

Provider-agnostic TypeScript toolkit for streaming, tool calls and chat UIs. Makers' products we track use it for AI SDK & agent framework (5).

What are the pros and cons of AI SDK?

Pros makers name: One TypeScript API across many model providers, so swapping models needs no code rewrite. Streaming and tool calling are dependable, with tight Next.js and React integration. A well-written core that other agent libraries build on, with version-matched docs shipped in the package. Cons: Easy to drop in but hard to make resilient: no resumable streaming after a dropped connection and no multi-device support. Streaming tool-call JSON is re-parsed on every chunk, which stalls on large payloads.

What are the alternatives to AI SDK?

For AI SDK & agent framework, the other options makers' products we track chose most are LangChain (66), Mastra (24), LlamaIndex (16), CrewAI (3), LangGraph (1). Open source or self-hostable among them: LangChain, Mastra, LlamaIndex, CrewAI, LangGraph. Each alternative, with the situation it fits, is on the AI SDK alternatives page.

Do developers switch to or away from AI SDK?

On GitHub, 4 public pull requests moved projects to AI SDK from tools we list, most from LangChain (4); 3 moved them away, most to LangChain (3).

When should you use AI SDK?

Your app is TypeScript and you want agents without adding a Python service (AI SDK & agent framework). Something else fits better if: you want a small typed agent loop in Python, close to the raw API, pick Pydantic AI or OpenAI Agents SDK; your agent works on files and runs commands, like a coding or research agent, pick Claude Agent SDK.

Does it work outside React and Next.js?

Yes. The UI hooks support React, Svelte, Vue.js, Angular and SolidJS, and AI SDK Core also runs in plain Node.js or Deno with no UI framework. source ↗

Can my backend be in Python?

Yes. useChat reads a documented stream protocol (Server-Sent Events with the x-vercel-ai-ui-message-stream: v1 header), so any backend that emits it works; the docs link a FastAPI example. source ↗

Do I need Vercel or the AI Gateway?

No. Plain model strings use Vercel AI Gateway as the default global provider, but you can set your own default, such as the OpenAI provider, or pass provider instances to call vendors directly. source ↗

Does it store chat history for me?

No. You write the load and save functions around useChat and streamText; the guide stores messages in files as an example and says to use a database in a real app. source ↗

Will upgrades break my code?

Breaking changes to stable APIs only come in major versions, which ship with migration guides and codemods where possible. APIs prefixed experimental_ can change in any release, so pin the exact version if you use them. source ↗

Can agents use tools from MCP servers?

Yes. The MCP client connects over HTTP (recommended for production), SSE or stdio, which is for local servers only, and gives the model the server's tools. source ↗

Is AI SDK free?

Yes — it is free, open-source software.

Is AI SDK open source or self-hostable?

Open source, and you can self-host it. source ↗

Can AI coding agents work with AI SDK?

No llms.txt, official MCP server or CLI found yet.

Who uses AI SDK?

5 makers' products we track, each with a source, and 241 open-source projects declare it in their code. The best known of them on Product Hunt: onlook (1,089 followers), Quartr (573 followers), Fuser (487 followers), Sparks AI (204 followers), Zyntax IDE (55 followers). source ↗