Show HN: Context Diet – trims tool output before it floods Claude Code's context

Category: devtools

Tags: ai-assistant, context-management, claude-code, log-summarization, developer-productivity

Score: 6.8/10 (Innovation: 6, Technical: 7, Documentation: 8, Utility: 6)

Context Diet is a Claude Code mod that intelligently trims large tool outputs (test logs, JSON, diffs, etc.) into concise digests before they flood the AI's context window, while preserving the full output in a file for later retrieval. It uses output-type-specific summarization rules and includes safety features like dry-run, command whitelisting, and self-monitoring of reopen rates. The project is notable for its targeted approach to a real pain point in AI-assisted development and for including empirical tests showing fewer tool calls and token savings.

Target audience: AI-assisted developers, particularly those using Claude Code for large codebases and verbose test/build outputs

Repository: https://github.com/drkokorev/context-diet · TypeScript · MIT · 8 stars

View on Hacker News