Skip to main content

Documentation Index

Fetch the complete documentation index at: https://motherfuckingsideproject.com/docs/llms.txt

Use this file to discover all available pages before exploring further.

​
Quick Start

Install AgentXP and serve your first markdown response in minutes.
1

Install the package

Add @reaganhsu/agentxp-next to your project.
npm install @reaganhsu/agentxp-next
2

Add middleware

Create middleware.ts in your project root (next to package.json). If your project uses a src/ directory, create src/middleware.ts instead.
middleware.ts
import { withAgentXP } from '@reaganhsu/agentxp-next'

export default withAgentXP()
That’s all you need for a working installation. AgentXP detects AI agents and serves markdown automatically from this point forward.Already have middleware? If you use Clerk, next-intl, or any other middleware, pass it as the second argument:
middleware.ts
import { withAgentXP } from '@reaganhsu/agentxp-next'
import { clerkMiddleware } from '@clerk/nextjs/server'

export default withAgentXP({}, clerkMiddleware())
AgentXP runs first. If the request is from an AI agent, it handles the response. Otherwise, it passes the request to your existing middleware unchanged.
3

Add the AgentExperience component (optional)

Add the AgentExperience component to your root layout to inject structured metadata for agents — a JSON-LD WebSite schema, a link to /llms.txt, and AI-specific robots directives.
import { AgentExperience } from '@reaganhsu/agentxp-next'

export default function RootLayout({ children }) {
  return (
    <html>
      <body>
        {children}
        <AgentExperience siteName="My Site" description="What my site does" />
      </body>
    </html>
  )
}
The component renders no visible UI. It injects:
  • JSON-LD structured data (WebSite schema)
  • A <link> to /llms.txt
  • AI-specific meta robots directives (noai, noimageai) based on your permissions settings
4

Verify it works

Start your dev server and send a request with Accept: text/markdown to any page:
curl -H "Accept: text/markdown" http://localhost:3000/
You should see clean markdown instead of HTML:
# Your Page Title

Your page content, converted to clean markdown.
Lists, links, tables, and code blocks are preserved.

---

*Source: [http://localhost:3000](http://localhost:3000)*
Check the response headers to confirm AgentXP is active:
curl -I -H "Accept: text/markdown" http://localhost:3000/

content-type: text/markdown; charset=utf-8
x-agentxp-agent: unknown
x-agentxp-version: 0.1.0
x-markdown-tokens: 342
content-signal: ai-train=no, ai-input=yes, search=yes
You can also verify the auto-generated endpoints:
# Site index for AI agents
curl http://localhost:3000/llms.txt

# Machine-readable capabilities manifest
curl http://localhost:3000/.well-known/agent-experience.json

​
Configuration

withAgentXP() accepts an optional configuration object. Here is a full example showing the available options: 
export default withAgentXP({
  transform: {
    contentSelector: 'main',
    excludeSelectors: ['.chat-widget', '.cookie-banner'],
    maxTokens: 8000,
  },
  permissions: {
    aiTrain: false,
    aiInput: true,
    search: true,
  },
  llmsTxt: {
    siteName: 'My Site',
    description: 'What my site does',
    pages: [
      { path: '/docs', title: 'Docs', section: 'Documentation' },
      { path: '/pricing', title: 'Pricing', section: 'Product' },
    ],
  },
})
See Configuration overview for the full reference.
AgentXP requires Next.js 14 or later.