LLMs.txt Integration

Overview

Docus automatically integrates nuxt-llms to prepare your documentation for Large Language Models (LLMs). This generates /llms.txt and /llms-full.txt files that provide AI-readable summaries and links to your documentation content.

All documentation pages are automatically indexed and included in the generated llms.txt files. No configuration required!

What is llms.txt?

The llms.txt file is a standardized format for exposing documentation to LLMs. It provides:

Site metadata - Title, description, and domain
Page index - List of all documentation pages with titles and descriptions
Direct links - URLs to each page for the LLM to fetch and read

This allows LLMs like Claude, GPT-4, and others to discover and understand your documentation structure.

View the live example: docus.dev/llms.txt

Generated Files

Docus generates two files:

/llms.txt

A compact version with essential information:

# My Documentation

A comprehensive guide to using our framework.

## Documentation

- [Getting Started](https://docs.example.com/getting-started): Learn how to install
- [Configuration](https://docs.example.com/configuration): Configure your site
- [Components](https://docs.example.com/components): Available UI components

/llms-full.txt

An extended version with additional metadata and more detailed descriptions. This is useful for LLMs that can handle larger context windows.

Default Configuration

Docus automatically extracts information from your package.json and deployment environment:

Field	Source	Example
`domain`	Deployment platform or `NUXT_SITE_URL` env var	`https://docs.example.com`
`title`	`package.json` name	`My Documentation`
`description`	`package.json` description	`A comprehensive guide...`
`full.title`	`package.json` name	`My Documentation (Full)`
`full.description`	`package.json` description	`Complete documentation...`

Custom Configuration

Override default values in nuxt.config.ts:

nuxt.config.ts

export default defineNuxtConfig({
  llms: {
    domain: 'https://docs.example.com',
    title: 'My Framework Documentation',
    description: 'A comprehensive guide to using My Framework',
    
    // Extended version configuration
    full: {
      title: 'My Framework Complete Documentation',
      description: 'The complete guide with examples, API reference, and best practices'
    }
  }
})

Raw Markdown Access

When nuxt-llms is enabled, Docus exposes a raw markdown endpoint that provides LLM-optimized content without HTML rendering. This reduces token usage and improves response speed for AI tools.

How It Works

Endpoint pattern: /raw/<content-path>.md

Use the same path as the page URL
Drop trailing /index
Keep the .md extension

Example mappings:

Page URL	Raw Markdown URL
`/en/getting-started`	`/raw/en/getting-started.md`
`/en/ai/assistant`	`/raw/en/ai/assistant.md`
`/configuration`	`/raw/configuration.md`

Features:

Content-Type: text/markdown; charset=utf-8
Auto-enrichment: Missing titles/descriptions are automatically prepended
LLMs.txt integration: Links in llms.txt point to raw markdown by default

Try it: Visit /raw/en/ai/llms-txt.md to see the raw markdown content.

Configuration

Customize raw markdown behavior:

nuxt.config.ts

export default defineNuxtConfig({
  llms: {
    contentRawMarkdown: {
      // Exclude specific collections from raw markdown exposure
      excludeCollections: ['landing', 'landing_en', 'landing_fr'],
      
      // Keep llms.txt links pointing to rendered HTML pages
      rewriteLLMSTxt: false
    }
  }
})

Disable Raw Markdown

To disable raw markdown access entirely:

nuxt.config.ts

export default defineNuxtConfig({
  llms: {
    contentRawMarkdown: false
  }
})

Markdown Redirection (Vercel)

This feature is currently only available when Docus is deployed on Vercel. It will become platform-agnostic once Nitro v3 supports global rewrites for multiple vendors.

When deployed on Vercel, Docus automatically configures intelligent routing to serve markdown content to AI agents and CLI tools.

Why?

AI agents like Claude Desktop use Accept: text/markdown headers by default. Serving raw markdown instead of rendered HTML saves significant data transfer and token costs.

Detection Methods

Docus detects AI agents and command-line tools using HTTP headers:

Accept header: Accept: text/markdown → redirect to markdown
User-agent: curl/.* → redirect to markdown

Redirect Rules

Original Request	Redirects To
`/`	`/llms.txt`
`/{locale}`	`/llms.txt`
`/{path}`	`/raw/{path}.md`

Example Usage

# Returns llms.txt
curl -H "Accept: text/markdown" https://docs.example.com/

Implementation Details

The markdown redirection is configured through Vercel’s routing system during build time. The markdown-rewrite module:

Detects Vercel deployment
Reads the generated llms.txt file
Extracts all documentation URLs
Creates Vercel routing rules
Writes rules to config.json

This happens automatically during the build process - no configuration required.

Advanced Configuration

Custom Content Filtering

Control which pages appear in llms.txt:

nuxt.config.ts

export default defineNuxtConfig({
  llms: {
    // Custom filtering function
    filter: (page) => {
      // Exclude draft pages
      if (page.draft) return false
      
      // Exclude private sections
      if (page.path.startsWith('/internal/')) return false
      
      return true
    }
  }
})

Custom URL Generation

Customize how URLs are generated in llms.txt:

nuxt.config.ts

export default defineNuxtConfig({
  llms: {
    // Use custom domain for links
    urlGenerator: (page, domain) => {
      return `${domain}${page.path}`
    }
  }
})

Multiple Language Support

For multi-language documentation, llms.txt automatically includes pages from all locales:

# My Documentation

## English Documentation

- [Getting Started](https://docs.example.com/en/getting-started)
- [Configuration](https://docs.example.com/en/configuration)

## French Documentation

- [Commencer](https://docs.example.com/fr/commencer)
- [Configuration](https://docs.example.com/fr/configuration)

Integration with AI Assistant

The llms.txt files work seamlessly with Docus’s built-in AI assistant:

Initial Discovery: When an AI tool connects to your docs, it reads llms.txt to understand structure
Raw Markdown Access: The assistant fetches pages via /raw/*.md endpoints for efficient processing
MCP Integration: The MCP server provides programmatic access to the same content

This creates a complete AI-ready documentation ecosystem.

Testing Generated Files

Verify your llms.txt files are generated correctly:

Build Your Site

Run the build command to generate static files:

npm run build

Check Output

Verify the files exist in your output directory:

ls .output/public/llms*.txt

View Content

Check the generated content:

cat .output/public/llms.txt

Test in Production

After deployment, visit your live URLs:

https://your-site.com/llms.txt
https://your-site.com/llms-full.txt

Best Practices

Descriptive Titles

Use clear, descriptive page titles that help LLMs understand content at a glance.

Comprehensive Descriptions

Write detailed page descriptions in frontmatter - these appear in llms.txt and help with discovery.

Logical Structure

Organize documentation hierarchically so LLMs can navigate relationships between topics.

Keep Content Updated

The llms.txt files regenerate on each build, ensuring AI tools always see current content.

Troubleshooting

llms.txt Not Generated

If llms.txt isn’t being generated:

Verify nuxt-llms is installed (it’s included by default in Docus)
Check that you have content in content/ directory
Run npm run build to generate static files
Check build output for any errors

Missing Pages

If some pages aren’t appearing:

Verify pages are in a content collection (not in public/)
Check that pages have frontmatter with title
Ensure pages aren’t excluded by custom filters
Rebuild the site after adding new pages

Raw Markdown 404 Errors

If raw markdown URLs return 404:

Verify contentRawMarkdown isn’t set to false
Check the path matches the content structure
Ensure the page exists and has been built
Try accessing the rendered HTML version first

nuxt-llms

Official nuxt-llms module documentation

MCP Server

Learn about the MCP server that complements llms.txt

AI Assistant

See how the assistant uses llms.txt for context

LLMs.txt Standard

Learn about the llms.txt format specification

Get Started

Concepts

Essentials

AI Features

​Overview

​What is llms.txt?

​Generated Files

​/llms.txt

​/llms-full.txt

​Default Configuration

​Custom Configuration

​Raw Markdown Access

​How It Works

​Configuration

​Disable Raw Markdown

​Markdown Redirection (Vercel)

​Why?

​Detection Methods

​Redirect Rules

​Example Usage

​Implementation Details

​Advanced Configuration

​Custom Content Filtering

​Custom URL Generation

​Multiple Language Support

​Integration with AI Assistant

​Testing Generated Files

​Best Practices

Descriptive Titles

Comprehensive Descriptions

Logical Structure

Keep Content Updated

​Troubleshooting

​llms.txt Not Generated

​Missing Pages

​Raw Markdown 404 Errors

​Further Reading

nuxt-llms

MCP Server

AI Assistant

LLMs.txt Standard

Build docs developers (and LLMs) love

Overview

What is llms.txt?

Generated Files

/llms.txt

/llms-full.txt

Default Configuration

Custom Configuration

Raw Markdown Access

How It Works

Configuration

Disable Raw Markdown

Markdown Redirection (Vercel)

Why?

Detection Methods

Redirect Rules

Example Usage

Implementation Details

Advanced Configuration

Custom Content Filtering

Custom URL Generation

Multiple Language Support

Integration with AI Assistant

Testing Generated Files

Best Practices

Troubleshooting

llms.txt Not Generated

Missing Pages

Raw Markdown 404 Errors

Further Reading