Skim

概要

This tool serves as an MCP Server for Claude Code, designed to combat the issue of AI agents being overwhelmed by large command outputs. When commands yield extensive data, such as API responses or log files, Skim intercepts these outputs, extracts only their essential schema or structure, and forwards this concise representation to the AI. The full data is temporarily saved for on-demand inspection, significantly reducing AI context window consumption, token costs, and preventing distraction by irrelevant details, thereby enhancing agent efficiency and focus.

主な機能

• Optimizes AI context by returning concise data structures instead of raw data.
• Saves full command outputs to local temporary files for later reference.
• Supports schema extraction for JSON, CSV, YAML, XML/HTML, and plain text logs.
• Automatically extracts schema from large command outputs.
• Provides dedicated tools for drilling into saved data, listing files, and cleanup.
• 3 GitHub stars

ユースケース

• Streamlining API development by quickly viewing response structures.
• Simplifying log analysis by previewing formats and drilling into specific entries.
• Efficiently analyzing database query results without loading all rows into context.