---
title: "Build AI-powered documents"
description: "Build AI agents that read and edit Tiptap documents in real time."
canonical_url: "https://tiptap.dev/docs/ai/ai-toolkit/overview"
---

# Build AI-powered documents

Build AI agents that read and edit Tiptap documents in real time.

- **1. Get access**

  Access AI Toolkit by purchasing the paid AI Toolkit add-on.
  Contact our team.
- **2. Install the package**

  The AI Toolkit package is open source. Install it from the public npm registry using npm or your
  preferred package manager.
- **3. Authenticate to the service**

  Configure a Tiptap Access Control JWT
  to access the AI Toolkit service.

Tiptap's **AI Toolkit** enables your AI agents to work with Tiptap documents inline. It includes tools you pass to your agents to
read and edit Tiptap documents, even without a browser-based editor.

> **Interactive demo:** [server ai agent chatbot](https://ai-toolkit-demos.vercel.app/server-ai-agent-chatbot)

### Use your own AI model & infrastructure

The AI Toolkit is framework-agnostic. It works with Vercel AI SDK, LangChain, OpenAI SDK, Anthropic SDK, or your own implementation.

### Built with efficiency in mind

The AI Toolkit adapts to cost-sensitive use cases. The available tools, like `tiptapEdit` and `proofread`, use an efficient format that allows AI to apply edits fast and keep token consumption low.

[Tiptap Shorthand](https://tiptap.dev/docs/ai/ai-toolkit/advanced-guides/tiptap-shorthand.md) (alpha) is a token-efficient format used for encoding Tiptap documents. It reduces AI token costs by up to 80% without sacrificing accuracy.

### Flexible deployment options

The AI Toolkit can be used as a cloud service, deployed on-premise on your infrastructure, or [embedded in your client-side codebase](https://tiptap.dev/docs/ai/ai-toolkit/client.md).