AI & LLM Integration

Complete AI and LLM integration following the official Fumadocs guide, making your documentation easily consumable by AI agents and large language models.

Overview

This site provides multiple ways for AI agents to access documentation:

🔎

Discovery /llms.txt endpoint lists all available docs

📄

Full Docs /llms-full.txt provides complete documentation

</>

Markdown .mdx and .md extensions for any page

✨

Smart Routing Automatic content negotiation

Features

LLM-Friendly Endpoints

`/llms.txt` - Discovery File

Standard discovery file for AI agents following the llms.txt specification.

curl https://yourdomain.com/llms.txt

Response:

# AI Web Feeds Documentation

> A collection of curated RSS/Atom feeds optimized for AI agents

## Documentation Pages

- [Getting Started](https://yourdomain.com/docs.mdx): Quick start guide
- [PDF Export](https://yourdomain.com/docs/features/pdf-export.mdx): Export docs as PDF
...

`/llms-full.txt` - Complete Documentation

All documentation in a single, structured text file optimized for RAG systems.

curl https://yourdomain.com/llms-full.txt

The format includes metadata header, table of contents, and structured page sections. See llms-full.txt Format for details.

Key Features:

Structured format with clear separators
Metadata header (date, page count, base URL)
Table of contents
Individual page sections with metadata
Optimized for AI parsing

Markdown Extensions

Access markdown source of any documentation page by appending .mdx or .md:

bash curl https://yourdomain.com/docs/getting-started.mdx Returns the markdown source of the page.

bash curl https://yourdomain.com/docs/getting-started.md Alternative markdown extension (same as .mdx).

bash curl -H "Accept: text/markdown" https://yourdomain.com/docs/getting-started Automatically serves markdown when AI agent requests it.

Content Negotiation

Middleware automatically detects AI agents and serves markdown content:

middleware.ts

import { isMarkdownPreferred } from "fumadocs-core/negotiation";

if (isMarkdownPreferred(request)) {
  // Serve markdown version
  return NextResponse.rewrite(new URL(`/llms.mdx${path}`, request.url));
}

When an AI agent sends Accept: text/markdown header, it automatically receives markdown content without changing the URL.

AI Page Actions

Interactive UI components on every documentation page:

Copy Markdown Button

One-click copy of page markdown to clipboard:

import { LLMCopyButton } from "@/components/page-actions";

<LLMCopyButton markdownUrl={`${page.url}.mdx`} />;

Features:

Client-side caching for performance
Loading state feedback
Success confirmation with checkmark

Dropdown menu with links to AI tools:

Open in GitHub - View source code
Open in Scira AI - Ask questions about the page
Open in Perplexity - Search with context
Open in ChatGPT - Analyze content

import { ViewOptions } from "@/components/page-actions";

<ViewOptions markdownUrl={`${page.url}.mdx`} githubUrl="https://github.com/..." />;

Implementation

File Structure

apps/web/
├── app/
│   ├── llms.txt/
│   │   └── route.ts              # Discovery endpoint
│   ├── llms-full.txt/
│   │   └── route.ts              # Full docs endpoint
│   ├── llms.mdx/
│   │   └── [[...slug]]/
│   │       └── route.ts          # .mdx handler
│   ├── llms.md/
│   │   └── [[...slug]]/
│   │       └── route.ts          # .md handler
│   └── docs/
│       └── [[...slug]]/
│           └── page.tsx          # With page actions
├── components/
│   └── page-actions.tsx          # AI UI components
├── middleware.ts                 # Content negotiation
└── next.config.mjs               # URL rewrites

Configuration

Source Config

Already configured in source.config.ts:

source.config.ts

export const docs = defineDocs({
  docs: {
    dir: "content/docs",
    includeProcessedMarkdown: true, // ✅ Required for LLM support
  },
});

Next.js Config

URL rewrites in next.config.mjs:

next.config.mjs

async rewrites() {
  return [
    {
      source: '/docs/:path*.mdx',
      destination: '/llms.mdx/:path*',
    },
    {
      source: '/docs/:path*.md',
      destination: '/llms.md/:path*',
    },
  ];
}

Usage

For AI Agents

bash # Discover all documentation curl https://yourdomain.com/llms.txt Returns a list of all available pages with descriptions.

bash # Get complete documentation curl https://yourdomain.com/llms-full.txt Returns all pages in a structured format.

bash # Get specific page as markdown curl https://yourdomain.com/docs/getting-started.mdx Returns markdown source of the page.

bash # Use content negotiation curl -H "Accept: text/markdown" https://yourdomain.com/docs/getting-started Automatically receives markdown content.

For Users

Copy Page as Markdown

Navigate to any documentation page
Click the Copy Markdown button
Paste into your AI tool or editor

Open in AI Tools

Click the View Options dropdown
Select your preferred AI tool:
- GitHub - View source code
- Scira AI - Ask questions
- Perplexity - Search with context
- ChatGPT - Analyze content

For Developers

Get LLM Text Programmatically

import { getLLMText, source } from "@/lib/source";

const page = source.getPage(["getting-started"]);
const markdown = await getLLMText(page);

Customize Page Actions

Edit components/page-actions.tsx to add more AI tools:

{
  title: 'Open in Claude',
  href: `https://claude.ai/new?content=${markdownUrl}`,
  icon: <ClaudeIcon />,
}

Update GitHub URLs

Edit app/docs/[[...slug]]/page.tsx:

githubUrl={`https://github.com/wyattowalsh/ai-web-feeds/blob/main/apps/web/content/docs/${page.file.path}`}

Performance

All endpoints are optimized for performance:

Endpoint	Caching Strategy	Generation
`/llms.txt`	`s-maxage=86400` (24h)	Dynamic
`/llms-full.txt`	`revalidate=false` (permanent)	Dynamic
`*.mdx` routes	`immutable`	Static
Middleware	Minimal overhead	Runtime
Copy button	Client-side cache	Client

Static generation ensures fast response times and minimal server load.

Benefits

For AI Agents

Easy discovery via /llms.txt
Complete context via /llms-full.txt
Granular access via .mdx extensions
Automatic detection via content negotiation
Optimized format for RAG systems

For Users

Quick markdown copy with one click
Direct AI tool links in View Options
Easy sharing with AI-friendly URLs
Better collaboration with AI assistants

For Developers

Standards-compliant following llms.txt spec
Performance-optimized with caching
Extensible architecture
Well-documented implementation

llms-full.txt Format - Detailed format specification
Testing Guide - Verify your integration
Quick Reference - Commands and endpoints

AI & LLM Integration

Overview

Features

LLM-Friendly Endpoints

`/llms.txt` - Discovery File

`/llms-full.txt` - Complete Documentation

Markdown Extensions

Content Negotiation

AI Page Actions

Copy Markdown Button

View Options Menu

Implementation

File Structure

Configuration

Source Config

Next.js Config

Usage

For AI Agents

For Users

Copy Page as Markdown

Open in AI Tools

For Developers

Get LLM Text Programmatically

Customize Page Actions

Update GitHub URLs

Performance

Benefits

For AI Agents

For Users

For Developers

External Resources

On this page