Tip: Lean Cortex MCP Server Delivers 60-99% Token Savings With 51 Tools and Cached Reads at ~13 Tokens

GitHub·medium signal

Lean Cortex (lean-ctx) is a Rust-based MCP server providing 51+ tools specifically designed to reduce token waste in AI coding agents by 60-99%. It works through two mechanisms: a shell hook that compresses CLI output before it reaches the LLM (60+ patterns), and an MCP server with cached file reads (~13 tokens per cached read), incremental deltas, dependency maps, cross-session memory, and multi-agent coordination. Compatible with Claude Code, Cursor, Copilot, Windsurf, and 23+ other agents without configuration. At 2,023 stars, it's a sleeper tool for anyone hitting context limits.