Gguf Mcp

Overview

Gguf Mcp is a Model Context Protocol server that inspects local GGUF and safetensors models, reading headers to report quantization, parameter counts, and VRAM requirements. Any MCP-compatible AI assistant can call its tools directly during conversation or agentic runs.

It runs entirely local via the npm package gguf-mcp, communicating with your MCP client over stdio. No authentication is needed. The current release is v0.1.1, licensed under MIT, with source code at arose26/gguf-mcp on GitHub.

Gguf Mcp sits in the AI & Agents category and works with Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Zed, and other MCP clients. If you maintain this server, claim your listing to earn the Verified badge and stay current.

Installation

// claude_desktop_config.json
{
  "mcpServers": {
    "gguf_mcp": {
      "command": "npx",
      "args": [
        "-y",
        "gguf-mcp"
      ]
    }
  }
}