Building Code Guardian Studio: An MCP Server for AI-Powered Code Refactoring

#mcp #claudeai #refactoring #typescript

Managing large codebases is challenging. Finding what to refactor first, tracking technical debt, and maintaining context across long coding sessions can be overwhelming.

I built Code Guardian Studio to solve these problems using the Model Context Protocol (MCP).

What is Code Guardian Studio?

It's an MCP server that provides 80+ tools for:

1. Code Analysis

Scan repositories for complexity metrics
Detect "hotspots" - files that need attention
Calculate Technical Debt Index (TDI)

2. Persistent Memory

Remember decisions across sessions
Store code patterns and conventions
Recall similar errors and their fixes

3. Code Guards

Validate code before commits
Catch fake tests, empty catch blocks
Security-focused rules

4. Workflow Management

Task tracking with priorities
Checkpoint system for progress
Process management for dev servers

Quick Start

# Install
npm install -g codeguardian-studio

# Add to Claude Code settings
ccg init

Example: Finding Refactoring Priorities

Claude: Use code_quick_analysis to scan the codebase

Result:
- Top hotspot: src/legacy/parser.ts (score: 8.5)
  - 500+ lines, deep nesting, high complexity
- Suggested: Split into smaller modules