Vision Squeezer
Reduces vision model token costs by snapping images to optimal tile boundaries for Claude, GPT-4o, GPT-5, and Gemini.
What it does
Reduces vision model token costs by snapping images to optimal tile boundaries for Claude, GPT-4o, GPT-5, and Gemini.
Vision Squeezer is a Rust-based MCP server that preprocesses images for LLM vision models to minimize token consumption and billing costs. It crops padding and resizes images to tile-boundary-aligned dimensions tailored to each model's tiling algorithm, supporting Claude, GPT-4o, GPT-5, and Gemini. Integrates as a background service with Claude Desktop, Cursor, VS Code Copilot, and JetBrains IDEs via the MCP protocol.
Capabilities
Server
Quality
deterministic score 0.55 from registry signals: · indexed on pulsemcp · has source repo · 1 github stars · registry-generated description present