AI

Use AI with SuperDoc: in coding agents, in your product, or in a visible editor.

SuperDoc does not bundle an AI model or send documents to an AI provider. It gives AI systems structured access to real .docx files (read, edit, comment, redline, save) through the same Document API the browser editor uses.

Pick your path

If you want to…	Use	Start here
Let Claude Code, Cursor, or Windsurf edit .docx files	MCP server	/ai/mcp/overview
Build AI document editing into your app	LLM tools + SDK	/ai/agents/llm-tools
Add AI actions to a visible editor	SuperDoc editor + Document API	/ai/agents/integrations

What AI can control

AI workflows can use SuperDoc to:

• Read document text, Markdown, HTML, and structure
• Search for clauses, paragraphs, comments, and tracked changes
• Insert, replace, and delete content
• Apply formatting
• Add and resolve comments
• Accept or reject tracked changes
• Save the result as a real .docx

The model never edits raw XML. It calls structured tools. SuperDoc applies the changes through the same document engine used by the editor.

I want my coding agent to edit DOCX files

Use the MCP server. Fastest path for Claude Code, Cursor, Windsurf, and other MCP clients.

claude mcp add superdoc -- npx @superdoc-dev/mcp

Then ask your agent to open a .docx file, inspect it, and make an edit.

MCP setup guide →

I want AI features inside my product

Use the SDK tool definitions. They work with OpenAI, Anthropic, Vercel AI SDK, and generic tool-calling loops.

npm install @superdoc-dev/sdk

import { createSuperDocClient, chooseTools, dispatchSuperDocTool } from '@superdoc-dev/sdk';

const client = createSuperDocClient();
await client.connect();

const doc = await client.open({ doc: './contract.docx' });
const { tools } = await chooseTools({ provider: 'openai' });

// Pass `tools` to your model.
// Dispatch tool calls with dispatchSuperDocTool(doc, name, args).

LLM tools guide →

I want AI output in a visible editor

Render the document in the browser editor for selection and visual feedback. Generate content with your model, then insert or replace it through the Document API.

const html = await callYourLLM(prompt);
editor.doc.insert({ target: cursor, value: html, type: 'html' });

Streaming editor pattern →

Data boundary

SuperDoc runs on your infrastructure. The editor, SDK, CLI, and MCP server do not send documents to SuperDoc Cloud.

If you connect an AI provider, your app or agent decides what document content to send. For sensitive documents, send the smallest useful excerpt instead of the full file.

Common patterns

Pattern	Recommended path
Draft a clause and insert it into the editor	Visible editor + editor.doc.insert
Review a contract and return redlines	SDK or MCP with tracked changes
Ask questions about a long document	Extract/search with Document API, send selected context to the model
Add comments instead of changing text	Comments tools (superdoc_comment or editor.doc.comments.*)
Build a review assistant inside your app	SDK LLM tools + your model provider
Let a coding agent fix a DOCX file locally	MCP server

What SuperDoc does not do

• It does not include an LLM.
• It does not automatically upload documents to an AI provider.
• It does not require a hosted SuperDoc service.
• It does not ask the model to rewrite raw DOCX XML.

SuperDoc gives the model safe document operations. Your app controls the model, prompts, permissions, and data flow.

Where to next

• AI Overview: the full AI section
• AI > MCP: Model Context Protocol setup
• AI > Agents > LLM tools: typed tool catalogs and integration patterns
• Document API: the operations all AI integrations share