DOCX MCP Server

Instructions

Implementation Reference

• src/index.ts:25-64 (handler)

  The handler function implementing the 'extract_text' tool. It resolves the DOCX file path, extracts raw text using mammoth.extractRawText, computes word count, structures the output as MCP content blocks, and handles errors appropriately.

  async ({ file_path }) => {
    try {
      const absolutePath = path.resolve(file_path)
  
      if (!fs.existsSync(absolutePath)) {
        throw new Error(`File not found: ${absolutePath}`)
      }
  
      const result = await mammoth.extractRawText({ path: absolutePath })
  
      return {
        content: [
          {
            type: 'text',
            text: JSON.stringify(
              {
                text: result.value,
                messages: result.messages,
                word_count: result.value
                  .split(/\s+/)
                  .filter((word: string) => word.length > 0).length,
              },
              null,
              2
            ),
          },
        ],
      }
    } catch (error) {
      return {
        content: [
          {
            type: 'text',
            text: `Error extracting text: ${(error as Error).message}`,
          },
        ],
        isError: true,
      }
    }
  }

• src/index.ts:22-24 (schema)

  Zod input schema for the 'extract_text' tool, defining the required 'file_path' parameter.

  {
    file_path: z.string().describe('Path to the .docx file'),
  },

• src/index.ts:19-65 (registration)

  MCP server registration of the 'extract_text' tool, specifying name, description, input schema, and handler function.

  server.tool(
    'extract_text',
    'Extract plain text content from a DOCX file',
    {
      file_path: z.string().describe('Path to the .docx file'),
    },
    async ({ file_path }) => {
      try {
        const absolutePath = path.resolve(file_path)
  
        if (!fs.existsSync(absolutePath)) {
          throw new Error(`File not found: ${absolutePath}`)
        }
  
        const result = await mammoth.extractRawText({ path: absolutePath })
  
        return {
          content: [
            {
              type: 'text',
              text: JSON.stringify(
                {
                  text: result.value,
                  messages: result.messages,
                  word_count: result.value
                    .split(/\s+/)
                    .filter((word: string) => word.length > 0).length,
                },
                null,
                2
              ),
            },
          ],
        }
      } catch (error) {
        return {
          content: [
            {
              type: 'text',
              text: `Error extracting text: ${(error as Error).message}`,
            },
          ],
          isError: true,
        }
      }
    }
  )