Document Extractor MCP Server

Instructions

Implementation Reference

• server.js:450-486 (handler)

  The 'extract_document' tool handler registers the tool, defines its input schema (zod), and executes the logic to extract and store content from Microsoft Learn or GitHub URLs.

  server.tool(
    'extract_document',
    'Extract document content from Microsoft Learn or GitHub URL and store in PocketBase',
    {
      url: z.string().url('Invalid URL format').describe('Microsoft Learn or GitHub URL to extract content from')
    },
    async ({ url }) => {
      try {
        await authenticateWhenNeeded();
        
        let docData;
        if (url.includes('learn.microsoft.com')) {
          docData = await extractFromMicrosoftLearn(url);
        } else if (url.includes('github.com') || url.includes('raw.githubusercontent.com')) {
          docData = await extractFromGitHub(url);
        } else {
          throw new Error('Unsupported URL. Only Microsoft Learn and GitHub URLs are supported.');
        }
        
        const record = await storeDocument(docData);
        
        return {
          content: [{
            type: 'text',
            text: `${record.isUpdate ? '🔄 Document updated' : '✅ Document extracted and stored'} successfully!\n\n` +
                  `**Title:** ${record.title}\n` +
                  `**ID:** ${record.id}\n` +
                  `**Source:** ${docData.metadata.source}\n` +
                  `**URL:** ${docData.metadata.url}\n` +
                  `**Word Count:** ${docData.metadata.wordCount}\n` +
                  `**Content Preview:** ${docData.content.substring(0, 200)}...`
          }]
        };
      } catch (error) {
        return toolErrorHandler(error);
      }
    }