Skip to main content
This guide shows you how to use WebHelp MCP Server to give AI tools instant access to any public Oxygen WebHelp documentation site.

​
Using the Live Demo

The fastest way to get started is using the hosted demo at webhelp-mcp.vercel.app.
1

Choose your WebHelp site

Identify the WebHelp documentation site you want to query. For this example, we’ll use the Oxygen XML Editor documentation:
https://www.oxygenxml.com/doc/versions/27.1/ug-editor/
Convert the URL to an endpoint path by removing https:// and the trailing slash:
www.oxygenxml.com/doc/versions/27.1/ug-editor
2

Construct your MCP endpoint

Append your endpoint path to the demo server:
https://webhelp-mcp.vercel.app/www.oxygenxml.com/doc/versions/27.1/ug-editor
This URL is your MCP server endpoint that AI tools can connect to.
3

Test the endpoint

You can verify the endpoint is working by visiting it in your browser. You should see an MCP server response.
The server dynamically loads the WebHelp search index on first request, so there may be a brief delay.

​
Configure AI Tools

Once you have your MCP endpoint, configure your AI tool to use it.

Claude Desktop

Add the MCP server to your Claude Desktop configuration

ChatGPT

Connect ChatGPT Deep Research to your documentation

Cursor

Configure Cursor’s AI features with MCP

Oxygen AI Positron

Use within Oxygen XML Editor’s AI assistant

​
Quick Configuration Example

For Claude Desktop, add this to your claude_desktop_config.json: 
{
  "mcpServers": {
    "oxygen-docs": {
      "url": "https://webhelp-mcp.vercel.app/www.oxygenxml.com/doc/versions/27.1/ug-editor"
    }
  }
}

​
Using the MCP Tools

Once connected, your AI tool has access to two powerful tools:

​
Search Tool

Search the documentation with semantic search and boolean operators. 
{
  "name": "search",
  "arguments": {
    "query": "wsdl"
  }
}
Search returns up to 10 results ranked by relevance. The server automatically uses semantic search when available and falls back to index-based search.

​
Fetch Tool

Retrieve the complete content of a document by its ID (from search results). 
{
  "name": "fetch",
  "arguments": {
    "id": "0:topics/wsdl-converter.html"
  }
}
The fetch tool automatically converts HTML content to clean Markdown, making it ideal for AI processing and analysis.
Query multiple documentation sites simultaneously by encoding multiple URLs.
1

Prepare your URLs

List the WebHelp sites you want to search together:
const sites = [
  "https://www.oxygenxml.com/doc/versions/27.1/ug-editor/",
  "https://www.oxygenxml.com/doc/versions/27.1/ug-author/"
];
2

Encode the URLs

Use the URL encoding utility (available in the source repository):
import { encodeUrls } from './lib/url-pack';
const encoded = encodeUrls(sites);
The encoding uses zlib compression and base64url encoding to create a compact representation.
3

Create federated endpoint

Use the federated prefix with your encoded URLs:
https://webhelp-mcp.vercel.app/federated/{encoded}
Search results will include documents from all specified sites.

​
Example Queries

Here are some example queries you can try:
Query: “XML validation schema”The search tool will find relevant documentation pages about XML validation features.
Query: “XSLT AND transformation”Boolean operators (AND, OR) help narrow down results to specific topics.
  1. First search: {"query": "content completion"}
  2. Get the ID from results: "0:topics/content-completion.html"
  3. Fetch full content: {"id": "0:topics/content-completion.html"}
The fetch tool returns the complete page content in Markdown format.

​
Next Steps

Deploy Your Own

Host your own instance on Vercel for production use

Semantic Search

Learn about intelligent search capabilities

Search API

Deep dive into the search tool implementation

Security

Understand security considerations for self-hosting
The demo server is for testing and development only. For production use, deploy your own instance to ensure reliability and security.

Build docs developers (and LLMs) love

Get started for freeTalk to us