mcp-hydrocoder-vision: Instructions file for Claude Code

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Commands

npm install      # Install dependencies
npm run build    # Build TypeScript to dist/ (optional, for production)
npm run dev      # Run in watch mode
npm start        # Start the server using tsx (development)

Architecture

This is an MCP (Model Context Protocol) server that enables image analysis using Qwen3 VL 4B running locally via LM Studio.

Key components:

• src/index.ts - Main server entry point using @modelcontextprotocol/sdk
• Three tools exposed: analyzeImage, extractText, describeForCode
• All image processing converts to base64 and sends to LM Studio's OpenAI-compatible API
• Input validation via Zod schemas
• Uses tsx for running TypeScript directly (no build required for development)

Environment variables:

• LM_STUDIO_URL - LM Studio API endpoint (default: http://localhost:1234/v1/chat/completions)
• VISION_MODEL - Model name (default: Qwen3-VL-4B-Instruct)

Configuration

To use with Claude Code (global installation via npm), add to ~/.claude/settings.json:

{
  "mcpServers": {
    "hydrocoder-vision": {
      "command": "npx",
      "args": ["-y", "mcp-hydrocoder-vision"],
      "env": {
        "LM_STUDIO_URL": "http://localhost:1234/v1/chat/completions",
        "VISION_MODEL": "Qwen3-VL-4B-Instruct"
      }
    }
  }
}

Requirements

• LM Studio must be running with Qwen3 VL 4B model loaded
• Node.js 18+
• TypeScript 5+ for development