Vision Squeezer

Overview

Vision Squeezer is a Model Context Protocol server that reduces vision-LLM token costs by snapping images to tile boundaries. It works with Claude, GPT, Gemini, Llama, Qwen and other vision models, allowing any MCP-compatible AI assistant to call its tools directly during conversations or agent runs.

The server runs locally via the npm package vision-squeezer, communicating over stdio on your machine. Install it with npx -y vision-squeezer. No authentication is required—it works immediately out of the box. The latest release is v0.6.1, with source code available at eralpozcan/vision-squeezer on GitHub.

Vision Squeezer belongs in the AI & Agents category and integrates with MCP clients including Claude Desktop, Claude Code, Cursor, VS Code, Windsurf, Zed and others. If you maintain this project, claim your listing to verify ownership and keep information up to date.

Installation

// claude_desktop_config.json
{
  "mcpServers": {
    "vision_squeezer": {
      "command": "npx",
      "args": [
        "-y",
        "vision-squeezer"
      ]
    }
  }
}