Skim

About

Skim is a production-ready Model Context Protocol (MCP) server that intelligently compresses code for large language model (LLM) context windows. Designed specifically to work with AI coding assistants like Claude Code, it automatically reduces code content by 60-95% while preserving structural integrity. This enables LLMs to analyze larger codebases without encountering 'context too long' limitations, leading to more accurate code suggestions and comprehensive project analysis.

Key Features

• Secure by Default with path validation, input sanitization, and rate limiting.
• DoS Protection through request and size limits, and timeout handling.
• 0 GitHub stars
• Production Monitoring capabilities including structured logging, health checks, and metrics.
• High Performance with fast execution via spawn and optimized buffering.
• Comprehensive Logging integrated with Winston for configurable levels.

Use Cases

• Analyzing large project architectures with AI coding assistants.
• Helping LLMs understand the core logic of large files by compressing them to essential structures.
• Facilitating detailed code reviews by enabling LLMs to process more context.