VitePress and Nextra: How to Make Your Docs Queryable by AI Agents

Docusaurus gets a lot of attention in the documentation tooling space, but two frameworks with equally large communities get far less: VitePress (the official Vue.js docs framework, used by Vue, Vite, Pinia, VueRouter, and hundreds of other projects) and Nextra (Next.js-native docs, used by Next.js itself, SWR, Turbo, and others).

If your docs run on either, adding a live MCP endpoint — so AI agents like Claude, Cursor, and Copilot can query your content with cited answers — takes under a minute and requires no changes to your framework configuration.

What AI agents need that static sites don't provide

VitePress and Nextra both generate clean HTML. They're fast. They're searchable. But when an AI agent tries to use your docs, it's not browsing — it's calling tools. The agent needs a structured interface: a way to ask a question and get a grounded answer with a source URL, not a list of pages to read.

Static HTML, even beautifully structured HTML, requires the agent to:

1. Fetch the page (which may fail on bot-blocked hosts)

2. Parse the HTML to extract the text

3. Synthesize an answer from multiple pages

4. Hope the content wasn't rendered by JavaScript after page load

VitePress and Nextra both use client-side routing in some configurations. An agent fetching https://vitepress.dev/guide/routing may get the app shell, not the content. AgentReady crawls and indexes at build time, so the agent always gets the actual text.

Adding MCP in 60 seconds

The setup is the same for both frameworks — AgentReady works at the URL level, not the framework level.

Step 1 — Index your site. Paste your docs URL at agentready.it.com. The crawler handles the rest. For VitePress sites this typically means your base URL (e.g. https://yourdocs.dev); for Nextra sites hosted on Vercel, it's the same.

Step 2 — Connect your AI client. Use the hosted MCP endpoint directly:

# Claude Desktop — add to claude_desktop_config.json
{
  "mcpServers": {
    "agentready": {
      "command": "npx",
      "args": ["@agentreadyweb/mcp"]
    }
  }
}

# Cursor — add to .cursor/mcp.json
{
  "mcpServers": {
    "agentready": {
      "url": "https://www.agentready.it.com/api/mcp"
    }
  }
}

Step 3 — Verify. Run a quick test from the CLI:

npx @agentreadyweb/mcp ask yourdocs.dev "How do I configure the sidebar?"
# Returns: cited answer with source URL from your actual docs

VitePress-specific notes

VitePress generates static HTML with a client-side router. The crawl works well on VitePress sites because the server-rendered HTML contains the full page text — the JavaScript enhances navigation but doesn't gate the content.

If your VitePress site has a robots.txt that blocks crawlers, you'll need to allow AgentReady's user agent or use the sitemap-based indexing mode. Most VitePress sites don't block crawlers by default.

VitePress also generates a sitemap.xml automatically if you enable it in defineConfig. AgentReady uses the sitemap when available to ensure full coverage:

// .vitepress/config.ts
export default defineConfig({
  sitemap: {
    hostname: 'https://yourdocs.dev'
  }
})

Nextra-specific notes

Nextra is a Next.js docs framework, which means it uses Next.js's static export or server-side rendering depending on how you've configured it. Both modes work with AgentReady.

One Nextra-specific quirk: some Nextra themes use next/font and dynamic CSS-in-JS that can confuse shallow crawlers. AgentReady's crawler uses full HTML extraction with readability parsing (similar to Firefox's reader mode), so it extracts the article text even if the surrounding layout is complex.

For Nextra sites, the most useful thing you can do after indexing is refresh on every deploy. Add this to your Vercel deploy hook or GitHub Action:

# .github/workflows/deploy.yml (add after your deploy step)
- name: Refresh AgentReady index
  run: npx @agentreadyweb/mcp refresh yourdocs.dev
  # Keeps the MCP index synchronized with your latest content

Adding llms.txt

If you want to go a step further, add an llms.txt to your site root. AgentReady generates one for you based on the indexed content — grab it from your site's detail page and host it at /llms.txt.

For VitePress: place the file in your public/ folder and it'll be served at the root.

For Nextra: place it in public/llms.txt — Next.js serves public files at the root automatically.

The upgrade path

Once indexed, your docs are queryable by any AI agent that connects to AgentReady's MCP. You don't need to maintain a plugin or a sidecar service. When you publish new content, refresh the index — that's the entire maintenance burden.

The next step, if you want to go deeper, is the agent analytics view: which questions are agents asking about your docs, and which ones aren't being answered confidently? That's the signal for what to document next.