Agentic AI
Code Analyzer
A multi-agent code analysis system built with LangGraph. Six specialized agents coordinate code review, security scanning, and performance optimization, then bring their findings together in one report.
Project Overview
A production-grade, multi-agent AI system that performs intelligent code analysis using specialized agents. The system leverages LangGraph for orchestrating a workflow of 6 specialized agents — each responsible for a different aspect of code quality — from preprocessing and routing to deep code review, security vulnerability scanning, and performance optimization.
The system automatically detects the programming language, calculates code complexity, runs security scans and Python linting via Ruff, then intelligently routes the analysis through the appropriate agent pipeline based on the code's characteristics. Results are aggregated into a comprehensive scored report that can be downloaded as Markdown or a styled PDF.
Key Features
Multi-Agent Architecture
6 specialized agents (Preprocessor, Router, Code Review, Security, Optimization, Aggregator) orchestrated via LangGraph with conditional routing.
Intelligent Routing
Automatically selects analysis depth (Quick / Deep / Security-Focused) based on code complexity and security scan results.
5 Functional Tools
Language detection, complexity scoring, security scanning, pattern checking, and Ruff linter — all executed during preprocessing.
PDF Report Export
Download styled code audit reports with scorecards, metrics tables, and prioritized recommendations as PDF documents.
Interactive Code Diff
Compare original vs. refactored code line-by-line with red (-) and green (+) annotations in a side-by-side view.
Git Repository Support
Clone public repositories via URL, index their source files, and select any file for multi-agent analysis directly.
Technologies Used
Architecture & Routing
Preprocessor Agent
Runs all 5 tools: language detection, complexity scoring, security scanning, pattern checking, and Ruff linting.
Router Agent
Analyzes tool results and decides analysis path: Quick, Deep, or Security-Focused routing.
Code Review Agent
Deep code review covering bugs, logic errors, readability, best practices, and refactored code suggestions.
Security Agent
Vulnerability analysis: injection, secrets exposure, OWASP Top 10, and secure code rewrites.
Routing Logic
The system intelligently routes code through different analysis pipelines:
- Quick: Simple code → Code Review → Aggregator
- Deep: Complexity ≥ 40 or 3+ functions → Code Review → Optimization → Aggregator
- Security-Focused: HIGH/CRITICAL security issues → Code Review → Security → Optimization → Aggregator