VisionSqueezer
Getting Started

MCP Setup

Wire the VisionSqueezer MCP server into Claude Code, Cursor, Zed, VS Code, Windsurf, and more.

VisionSqueezer ships an MCP (Model Context Protocol) server that exposes image optimization as a tool your AI agent can call directly. Thanks to npx -y, no global install is required.

Quick install (Claude Code)

Terminal
claude mcp add vision-squeezer -- npx -y vision-squeezer

Editor configurations

Use the interactive selector on the home page to copy the exact config for your editor, or refer to the snippets below.

Add to .cursor/mcp.json:

{
  "mcpServers": {
    "vision-squeezer": {
      "command": "npx",
      "args": ["-y", "vision-squeezer"]
    }
  }
}

MCP tools

Once registered, the server exposes three tools:

ToolPurposeKey args
optimize_imageStandard optimizationimage_base64, target_model (claude / gpt4o / gpt5 / gemini), mode (standard / ocr / auto), output_format (jpeg / webp / avif)
sandbox_executeThink-in-Code: run atomic ops locallyoperations[]crop, grayscale, binarize, resize, contrast, brightness
get_savings_statsCumulative token / USD savings
Pin the version (npx -y vision-squeezer@X.Y.Z) in shared configs. An unpinned npx cache can freeze users on whatever tarball resolved first.

Verify

Terminal
npx -y vision-squeezer --version

If the MCP server fails to connect, run vision-doctor (Claude Code skill) — it probes the registered command and surfaces the root cause.