MCPFast / Tools / Squeez: Hook-based token compressor for AI CLIs

GitHubTool★★★★☆

Squeez: Hook-based token compressor for AI CLIs

Squeez compresses tokens for 5 AI CLIs, offering up to 95% reduction, cross-call dedup, and an MCP server.

View on GitHub

Squeez: Hook-based Token Compressor for AI CLIs

Squeez is a utility designed to significantly reduce token consumption for AI Command Line Interface (CLI) applications. By employing a hook-based approach, Squeez integrates with existing CLIs to compress input tokens before they are sent to AI models. This results in substantial cost savings and improved performance for developers working with AI-powered command-line tools. It offers advanced deduplication and can leverage an MCP server for enhanced functionality.

What Squeez Does

Squeez acts as an intermediary layer between your AI CLI and the AI model. It intercepts the tokens generated by the CLI, analyzes them for redundancy, and compresses them using various techniques. This compression can achieve reductions of up to 95%, meaning you send far fewer tokens to the AI model for processing. This is particularly beneficial for repetitive or verbose inputs common in development workflows. Squeez also provides cross-call deduplication, ensuring that identical or highly similar token sequences are not processed multiple times across different invocations.

Key Features

Who Squeez is For

Squeez is an essential tool for AI developers who frequently use AI-powered CLIs. This includes individuals and teams building or integrating AI into their command-line workflows, such as those working with code generation tools, AI-assisted debugging, or automated scripting. If you are concerned about the cost of AI API calls, looking to improve the speed of your AI CLI interactions, or dealing with large amounts of repetitive input, Squeez offers a practical and effective solution.