MCP Compress

What ShipItAndPray/mcp-compress MCP server does

ShipItAndPray/mcp-compress MCP server exposes seven tools for handling compressed text data through the Model Context Protocol. Agents can compress text, JSON, or CSV; restore the original content; inspect how suitable data is for compression; and maintain a compressed key-value store on disk.

The compress tool can select an algorithm automatically from gzip, Brotli, and deflate. Its result includes base64-encoded compressed content and measurements such as the compression ratio and percentage saved. The operation is lossless: decompress is intended to reproduce the original data byte-for-byte.

Analysis is available before compression. The analyze tool reports Shannon entropy, a compressibility rating, comparisons across the supported algorithms, and a recommendation. Storage operations use a caller-provided key. store compresses and saves data, retrieve restores it, list reports stored entries and their sizes, and stats summarizes item counts, saved bytes, and the overall ratio.

How it works

With automatic selection, ShipItAndPray/mcp-compress MCP server tests gzip, Brotli, and deflate against the supplied data and chooses the smallest result. The README describes Brotli as the usual winner for text, but the selection process is data-dependent. The server stores compressed key-value data under ~/.mcp-compress/, allowing later retrieval by key.

The MCP implementation communicates over standard input and output, so an MCP client can launch it as a local process. The repository describes the implementation as pure Node.js with zero dependencies. Reported examples include stronger compression for repetitive Markdown, configuration, SQL results, logs, JSON responses, time-series data, and CSV, though those figures are benchmark results for the listed samples rather than guarantees for every input.

Setup and configuration

Install and run ShipItAndPray/mcp-compress MCP server with the npm package through npx:

npx -y mcp-compress

For a client configuration, use npx as the command and pass -y followed by mcp-compress as the arguments. The README also documents running node index.js after cloning the repository and starting it from the project directory. No API key, service account, or environment variable is described as required.

The documented transport is MCP over stdio. The repository specifically provides a Claude Code configuration and states that the process can be used with any MCP-compatible AI agent.

Tools and capabilities

• compress: Compresses text, JSON, or CSV and returns base64 data, the selected algorithm, and ratio information.
• decompress: Restores compressed content and supports lossless round trips.
• analyze: Compares algorithms and reports entropy, compressibility, and a recommendation.
• store: Compresses data and persists it under a key.
• retrieve: Returns data saved under a key.
• list: Shows stored items, sizes, and compression ratios.
• stats: Reports aggregate item counts, bytes saved, and overall compression ratio.

Limitations and notes

Compression results depend on the input. The README’s ratios come from specific benchmark samples and should not be treated as fixed performance expectations. The documented data examples focus on text-oriented content, including JSON and CSV. Storage persistence is local to the machine running the process and uses the documented ~/.mcp-compress/ location. The supplied material does not describe remote hosting, authentication, access controls, encryption, or configurable storage paths.