aicodeking/leanrea

What it solves

leanrea provides a lightweight alternative to complex reverse-engineering agent setups. It eliminates the need for heavy MCP (Model Context Protocol) servers and hundreds of pre-defined tools, which can bloat context windows, increase latency, and drive up costs for AI agents performing reverse-engineering tasks.

How it works

Instead of wrapping tools in an MCP layer, leanrea provides a single 8 KB "skill" file (a set of instructions and best practices) that guides an AI agent to use standard Bash commands and native system utilities directly. It teaches the agent how to identify file types, select the fastest command for a specific target (such as Electron apps, Mach-O binaries, or Go binaries), and execute efficient reconnaissance patterns without the overhead of external tool definitions.

Who it’s for

This project is for developers and security researchers who use AI agents (like Claude Code, Cursor, or Gemini CLI) to perform reverse engineering and want to optimize their agent's performance, reduce token usage, and minimize execution time.

Highlights

• Significant Efficiency: Reduces token usage by up to 60% and wall time by 38% compared to traditional MCP-based reverse-engineering tools.
• Minimalist Design: Replaces a 25 MB package and hundreds of tool definitions with a single 8 KB text file.
• Broad Compatibility: Works with any agent that supports custom skills or system prompts, including Claude Code, Cursor, and others.
• Optimized Patterns: Includes expert-level heuristics for handling complex targets like minified web bundles, Electron apps, and native binaries.