Elenchus

关于

Elenchus is a Model Context Protocol (MCP) server that implements an adversarial approach to code verification. Unlike traditional static analysis, it orchestrates a dynamic debate between Verifier and Critic AI agents, systematically questioning and refuting findings to expose contradictions and reach a robust understanding of code quality. This dialectical reasoning system aims for intent-based semantic analysis, adaptive convergence across five critical categories (Security, Correctness, Reliability, Maintainability, Performance), and provides explicit negative assertions on clean code, going beyond mere checklist-based evaluation.

主要功能

• Adversarial Debate System: Orchestrates a multi-round Verifier↔Critic loop for systematic issue discovery and challenge.
• Intent-Based Convergence: Utilizes semantic understanding across 5 key categories to achieve adaptive verification convergence.
• 4 GitHub stars
• LLM-Based Evaluation (Optional): Leverages Large Language Models for advanced assessments, including convergence quality, severity classification, and false positive detection.
• Automatic Impact Analysis: Builds multi-language dependency graphs to predict ripple effects, calculate cascade depth, and assess risk.
• Multi-Language Support: Provides dependency analysis for 15+ languages via tree-sitter AST parsing (e.g., TypeScript, Rust, Python).

使用案例

• Conducting detailed code reviews for correctness, reliability, and maintainability issues across a codebase.
• Performing comprehensive security audits for specific code modules or directories.
• Identifying and validating edge cases within critical code sections, such as payment processing logic.