Dokuwiki MCP Server

An MCP server for interacting with Dokuwiki API, allowing LLMs to find and retrieve content from Dokuwiki pages. The server provides AI-powered semantic similarity matching as the preferred method for finding relevant content, with traditional search as a fallback option.

Requirements

• Node.js 16 or higher
• A DokuWiki instance with:
  • JSON-RPC API enabled
  • Core search functionality
  • AI Chat plugin for similar pages functionality (strongly recommended for optimal results)

Setup

1. Install dependencies:

   npm install
   

2. Configure API access:

   export DOKUWIKI_API_URL="https://your-dokuwiki-instance/lib/exe/jsonrpc.php"
   export DOKUWIKI_API_TOKEN="your-api-token"
   

   Alternatively, edit the src/config.ts file directly.

   The configuration includes:

   • baseUrl: The URL to your DokuWiki JSON-RPC API endpoint
   • token: Your API authentication token (if required)
   • endpoints: The API methods to call for each function:
     • searchPages: "/core.searchPages"
     • getPage: "/core.getPage"
     • similar: "/plugin.aichat.similar" (requires the AI Chat plugin)

3. Build the project:

   npm run build
   

4. Run the server:

   npm start
   

Available Tools

getSimilarDokuwikiPages (Preferred)

Find pages in Dokuwiki that are semantically similar to a given query. This tool uses AI-powered similarity matching and should be your first choice when looking for relevant content.

Parameters:

• query: Query to find similar Dokuwiki pages for (string)

Returns:

• A formatted list of similar pages with titles and IDs

searchDokuwikiPages (Fallback)

Search for pages in Dokuwiki based on a query string. Use this tool only when getSimilarDokuwikiPages returns no results or when you need to search for specific terms rather than concepts.

Parameters:

• query: Search query for Dokuwiki pages (string)

Returns:

• A formatted list of matching pages with titles, IDs, and snippets showing where the search terms were found

getDokuwikiPage

Retrieve the content of a specific Dokuwiki page.

Parameters:

• pageId: ID of the Dokuwiki page to retrieve in a format like 'namespace:page_name' (string)

Returns:

• The full content of the requested page with a title derived from the page ID

Using with MCP Inspector

To test with the MCP Inspector, run:

npx @modelcontextprotocol/inspector node dist/index.js

Data Models

The server uses the following data models for interacting with the DokuWiki API:

Search Results

interface DokuwikiSearchResult {
  score: number;      // The number of hits this result got
  snippet: string;    // The HTML formatted snippet showing where the search term was found
  id: string;         // The page ID
  revision: number;   // The page revision aka last modified timestamp
  size: number;       // The page size in bytes
  title: string;      // The page title
  permission: number; // The current user's permissions for this page
}

Page Content

interface DokuwikiPageContent {
  text: string;       // The page content
}

Page Information

interface DokuwikiPageInfo {
  id: string;         // The page ID
  revision: number;   // The page revision aka last modified timestamp
  size: number;       // The page size in bytes
  title: string;      // The page title
  permission: number; // The current user's permissions for this page
  hash?: string;      // MD5 sum over the page's content (if available)
  author?: string;    // The author of this page revision (if available)
}

Similar Pages

interface DokuwikiSimilarResult {
  page: string;       // The page path or namespace:name format
  title: string;      // The human-readable title of the page
  score: number;      // Similarity score indicating relevance (higher is more relevant)
  lang: string;       // Language code of the page content
}

Response Format

All API responses follow this standard format:

interface DokuwikiApiResponse<T> {
  result: T;          // The result data, type varies by endpoint
  error: {
    code: number;     // Error code (0 indicates success)
    message: string;  // Error message or "Success" when successful
  };
}

Usage with LLM Integrations

This server implements the Model Context Protocol (MCP) and can be used with any LLM application that supports MCP.

Example:

import { Client } from "@modelcontextprotocol/sdk/client/index.js";
import { StdioClientTransport } from "@modelcontextprotocol/sdk/client/stdio.js";

const transport = new StdioClientTransport({
  command: "node",
  args: ["dist/index.js"]
});

const client = new Client({ name: "example-client", version: "1.0.0" });
await client.connect(transport);

// Example 1: Find similar pages (preferred approach)
const similarResult = await client.callTool({
  name: "getSimilarDokuwikiPages",
  arguments: {
    query: "example topic"
  }
});
console.log("Similar pages:", similarResult.content[0].text);

// Example 2: Fallback to search if no similar pages found
if (similarResult.content[0].text.includes("No similar pages found")) {
  const searchResult = await client.callTool({
    name: "searchDokuwikiPages",
    arguments: {
      query: "example topic"
    }
  });
  console.log("Search results:", searchResult.content[0].text);
}

// Example 3: Get a specific page
const pageResult = await client.callTool({
  name: "getDokuwikiPage",
  arguments: {
    pageId: "namespace:page_name"
  }
});
console.log("Page content:", pageResult.content[0].text);