Multimodel MCP

🚀 GLM-4.5V MCP Server

The MCP server with multi-modal capabilities of GLM-4.5V, providing image processing, visual query, and file processing functions.

🚀 Quick Start

The GLM-4.5V MCP Server offers a range of functionalities for image and file processing. Follow the steps below to get started.

✨ Features

• read_image: Read local/URL images and return dataURL and size information.
• vision_query: Invoke GLM-4.5V to perform OCR/QA/detection on images.
• process_file: Use GLM-4.5V to process files (upload and extract content).

📦 Installation

npm install

📦 Configuration

Copy the environment variable file:

cp .env.example .env

Edit the .env file and fill in your GLM API Key:

GLM_API_KEY=your_api_key_here
GLM_BASE_URL=https://open.bigmodel.cn/api/paas/v4/chat/completions

📦 Build

npm run build

📦 Run

npm start

Or use the startup script:

./start-mcp.sh

📦 Development Mode

npm run dev

📦 Test Functionality

Create a test file and verify the functionality:

node scripts/test-file-processing.js

📚 Documentation

read_image

Read an image and return a base64-encoded dataURL.

Parameters:

• path (Required): Image path or URL.
• maxSide (Optional): Maximum side length for scaling.

vision_query

Analyze an image using GLM-4.5V.

Parameters:

• path (Required): Image path or URL.
• prompt (Required): Query prompt.
• mode (Optional): Query mode ("describe", "ocr", "qa", "detect").
• returnJson (Optional): Whether to return the result in JSON format.

process_file

Process files using GLM-4.5V, supporting file uploads and content extraction in multiple formats.

Supported File Formats:

• Documents: PDF, DOCX, DOC, XLS, XLSX, PPT, PPTX, CSV, TXT
• Images: PNG, JPG, JPEG

File Size Limits:

• Image files: Maximum 5MB
• Other files: Maximum 50MB

Parameters:

• filePath (Required): Local file path.
• extractPrompt (Optional): Content extraction prompt to guide how to extract file content.

Return Result:

{
  "ok": true,
  "fileId": "file-xxx",
  "content": "Extracted file content...",
  "fileType": "PDF Document",
  "filename": "document.pdf",
  "metadata": {
    "uploadTime": 1234567890,
    "fileSize": 1024000,
    "processingTime": 5000
  }
}

💻 Usage Examples

Process a PDF Document

# Invoke via MCP
{
  "tool": "process_file",
  "arguments": {
    "filePath": "./documents/report.pdf",
    "extractPrompt": "Please extract the main content and key information from the document."
  }
}

Process an Excel Spreadsheet

{
  "tool": "process_file",
  "arguments": {
    "filePath": "./data/sales.xlsx",
    "extractPrompt": "Please analyze the spreadsheet data and summarize the sales trends."
  }
}