GLM MCP Ultimate Server

The GLM MCP Ultimate server allows Claude Code and other MCP-compatible clients to utilize GLM (Zhipu AI) models. It provides an autonomous coding agent, chat capabilities, image analysis, and text embeddings.

Features

• Focused Task Execution: Run GLM as a sub-agent to autonomously complete specific tasks within a fresh, isolated context before returning findings to the orchestrator.
• GLM Chat: Send direct prompts to GLM models for quick questions or second opinions.
• GLM Vision: Analyze images using GLM-4V by providing a URL or base64 data.
• GLM Embeddings: Generate high-quality text embeddings for search and retrieval tasks.
• SQLite Persistence: Robust storage for sessions, messages, and agent steps.

Prerequisites

• Node.js 18 or later
• A GLM API key from Z.ai - Recommended: GLM Coding Plan (10% off with this link)

Installation

Option 1: Easy Installation (NPM)

Install the package globally:

npm install -g @glitterkill/glm-mcp-ultimate

Option 2: Manual Installation (Development)

1. Clone the repository:

   git clone https://github.com/your-repo/glm-mcp-ultimate.git
   cd glm-mcp-ultimate

2. Install dependencies and build:

   npm install
   npm run build

Configuration

Using with Claude Desktop

To use this server with Claude Desktop, add it to your configuration file:

• Windows: %APPDATA%\Claude\claude_desktop_config.json
• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json

Add the following entry to the mcpServers object:

If installed via NPM:

{
  "mcpServers": {
    "glm": {
      "command": "npx",
      "args": ["-y", "@glitterkill/glm-mcp-ultimate"],
      "env": {
        "GLM_API_KEY": "your_api_key_here"
      }
    }
  }
}

If installed manually:

{
  "mcpServers": {
    "glm": {
      "command": "node",
      "args": ["/path/to/glm-mcp-ultimate/dist/index.js"],
      "env": {
        "GLM_API_KEY": "your_api_key_here"
      }
    }
  }
}

Available Tools

Autonomous Agent

• glm_run_task: Execute a single, focused task using a fresh GLM session. This tool minimizes context pollution by spawning a new session per task, operating autonomously up to a specified step limit, and returning findings directly to the orchestrating LLM.

General Tools

• glm_chat: Send a prompt to a GLM model and receive a text response.
• glm_vision: Provide an image and a prompt for visual analysis using GLM-4V.
• glm_embeddings: Convert text into vector embeddings using the GLM embedding model.

Agent Capabilities

When running in autonomous mode, the GLM agent can perform the following actions:

• File Operations: Read, create, and edit files using precise text replacement.
• System Commands: Execute shell commands to run tests, install packages, or perform version control.
• Code Exploration: List files and search for patterns across the codebase.
• Task completion: Provide a detailed summary once the objective is reached.

Environment Variables

• GLM_API_KEY: Required. Your Zhipu AI API key (Z.ai Coding Plan keys supported).
• GLM_MODEL: Optional. The default GLM model to use (defaults to glm-5).
• GLM_API_BASE: Optional. Override the base URL for the GLM API (defaults to https://api.z.ai/api/coding/paas/v4).

CLI Command Usage

In addition to acting as an MCP server, GLM MCP Ultimate provides direct CLI commands for terminal users. This allows you to interact with the models without needing an MCP client like Claude Code.

Global Installation: If you installed the package globally, you can use the glm-mcp-ultimate command directly:

# General Chat
glm-mcp-ultimate chat "Write a python script to reverse a string"

# Start Autonomous Agent
glm-mcp-ultimate agent "Refactor the authentication module" -d ./src/auth

# Image Analysis
glm-mcp-ultimate vision "Describe this image" https://example.com/image.jpg

# Generate Embeddings
glm-mcp-ultimate embeddings "Hello world"

Using NPX: If you prefer not to install globally, you can use npx:

npx -y @glitterkill/glm-mcp-ultimate chat "What is the capital of France?"

Note: Ensure your GLM_API_KEY environment variable is set before running these commands.