Skip to content
denote.sh

How Denote makes your documentation a first-class data source for AI agents

AI Native

See also: The API Reference documents the specific endpoints (JSON API, llms.txt) mentioned on this page.

Every Denote site is designed to serve two audiences: humans browsing your docs, and AI agents consuming your knowledge programmatically.

Why It Matters

AI coding assistants like Cursor, Claude, and ChatGPT are now the primary way many developers interact with documentation. When someone asks "how do I configure X?", the AI needs to find and read your docs — fast, accurately, and in a format it understands.

Most documentation frameworks treat AI access as an afterthought. Denote builds it in from day one.

Three Layers of AI Access

1. llms.txt — AI Discovery

Every Denote site auto-generates /llms.txt following the llms.txt standard. This is a curated index of your documentation that tells AI agents what's available and where to find it.

text
# My Project

> Project documentation

## Docs

- [Introduction](https://docs.example.com/docs/introduction): Getting started guide
- [API Reference](https://docs.example.com/docs/api): Complete API documentation

There's also /llms-full.txt — your entire documentation concatenated into a single file, optimized for context windows.

2. JSON API — Structured Data

The /api/docs endpoint serves your documentation as structured JSON, ready for RAG pipelines, embedding generation, or direct context injection:

json
{
  "name": "My Project",
  "pages": [
    {
      "slug": "introduction",
      "title": "Introduction",
      "description": "Getting started guide",
      "content": "# Introduction\n\nWelcome to...",
      "headings": [
        { "level": 1, "title": "Introduction", "id": "introduction" }
      ]
    }
  ]
}

3. MCP Server — Live Tool Access

The built-in MCP server turns your docs into tools that AI agents can search and read on demand.

Enable it in your config:

typescript
// denote.config.ts
export const config = {
  // ... other config
  ai: {
    mcp: true,
  },
};

MCP on Your Deployed Site

When mcp: true is set, your deployed docs site itself becomes an MCP endpoint at /mcp. No local setup needed — AI agents connect directly to your live docs URL.

This is the key differentiator: your documentation site isn't just web pages, it's a live API that any MCP-compatible tool can query.

Point any MCP client to your site's /mcp endpoint:

json
{
  "mcpServers": {
    "my-docs": {
      "url": "https://docs.example.com/mcp"
    }
  }
}

The endpoint supports Streamable HTTP with full CORS, so it works from any origin.

Local MCP

During local development, point your MCP client at your dev server's /mcp endpoint:

json
{
  "mcpServers": {
    "my-docs": {
      "url": "http://localhost:8000/mcp"
    }
  }
}

Available Tools and Resources

The MCP server exposes tools for searching, reading, and dumping your docs, plus resources for listing and accessing individual pages. See the API Reference — MCP Server for the full list of tools and resources.

Content Tags

Use frontmatter fields to improve AI understanding of your pages:

yaml
---
title: Authentication
ai-summary: OAuth2 and API key auth with JWT tokens, refresh flows, and scoped permissions.
ai-keywords:
  - authentication
  - oauth2
  - jwt
---

See Writing Content for full details on AI content tags.

Zero Configuration

All AI endpoints are generated automatically from your Markdown files. No plugins, no API keys, no indexing pipeline. Write your docs, and the AI layer just works.

Next Steps

Edit this page →