Specialized Code Review Tools for Copilot

A Model Context Protocol (MCP) server that provides specialized code review capabilities using established critic methodologies and fixing strategies.

Example Usage

The workflow has two steps in your VS Code agent chat:

1. Analysis: Use a critic tool to analyze code and identify issues
2. Fix: Apply a fixer strategy to address the identified problems

Step 1: Analyze Code

Start a new chat to clear context, then write a request like:

review-design this React component
review-c-memory myfunc()
review-sql-security app_schema.sql

Step 2: Apply Fixes

Choose one:

fix-conservative
fix-comment
fix-zealot

Features

• 36+ Specialized Critics: Each critic focuses on specific aspects of code quality
• 3 Fixing Strategies: Methods for addressing critic findings
• Explicit Invocation: Users explicitly choose which critic or fixer to apply to their code
• Established Methodologies: Uses frameworks from the ai-review repository
• Two-Step Workflow: Analysis first, then apply fixing strategy

Setup

Prerequisites

• Node.js 18.0 or higher
• VS Code with GitHub Copilot extension

Installation

1. Clone with submodules to get critic frameworks:

   git clone --recursive https://github.com/jonels-msft/mcp-review.git
   cd mcp-review
   
   # OR if already cloned:
   # git submodule update --init --recursive
   

2. Install dependencies:

   npm install
   

VS Code Configuration

Add the MCP server to your VS Code settings:

1. Open VS Code settings (JSON):

   • Press Ctrl+Shift+P (Windows/Linux) or Cmd+Shift+P (Mac) to open the command palette
   • Type "Preferences: Open User Settings (JSON)" and select it
   • This opens your settings.json file

2. Add the following to the mcp.servers section (create the section if it doesn't exist):

{
    "mcp": {
        "servers": {
            "mcp-review": {
                "command": "node",
                "args": [
                    "/path/to/your/mcp-review/src/index.js"
                ]
            }
        }
    }
}

{
    "mcp": {
        "servers": {
            "mcp-review": {
                "command": "wsl",
                "args": [
                    "node",
                    "/path/to/your/mcp-review/src/index.js"
                ]
            }
        }
    }
}

{
    "mcp": {
        "servers": {
            "mcp-review": {
                "command": "wsl",
                "args": [
                    "/home/username/.nvm/versions/node/v18.20.8/bin/node",
                    "/path/to/your/mcp-review/src/index.js"
                ]
            }
        }
    }
}

Important: Replace /path/to/your/mcp-review/ with the actual path where you cloned this repository (use Unix-style paths for WSL).

After adding the configuration, restart VS Code. VS Code will invoke the server automatically when you use the review tools in Copilot chat.

Available MCP Tools

The server loads all critic methodologies from the ai-review/critic directory and fixer strategies from the ai-review/fixer directory.

Critic MCP Tools

All critics follow the naming pattern review-<critic-name> and are organized by category:

Design & UI Critics

• review-design - UI and user experience review
• review-color - Color usage and accessibility
• review-data-visualization - Charts and data presentation

Security Critics

• review-sql-security - SQL injection and database security
• review-sql-data-integrity - Data validation and constraints

Performance Critics

• review-algorithm-performance - Algorithm efficiency analysis
• review-algorithm-correctness - Algorithm correctness verification
• review-logging-performance - Logging overhead analysis

Memory & Resource Management

• review-c-memory - C memory management review
• review-c-error-handling - C error handling patterns
• review-c-portability - Cross-platform C code

Error Handling

• review-general-error-handling - Error handling best practices
• review-procedural-error - Procedural error management

Data & Logic

• review-procedural-data - Data structure and flow analysis
• review-procedural-functions - Function design and organization
• review-procedural-flow - Control flow analysis
• review-floating-point - Numerical precision and floating-point issues
• review-datetime - Date/time handling correctness
• review-unicode - Unicode and internationalization

Standards & Compliance

• review-sql-standard-compliance - SQL standard adherence
• review-posix - POSIX compliance and portability
• review-unix-interface - Unix philosophy and interfaces

Development Practices

• review-logging-practices - Logging strategy and implementation
• review-terminology - Naming conventions and clarity
• review-writing - Documentation and comments
• review-estimation - Complexity and effort estimation
• review-benchmarking - Performance measurement practices

Specialized Critics

• review-meson - Meson build system review
• review-probability - Statistical and probabilistic code
• review-relational - Database design and relationships
• review-prompting - AI prompt engineering
• review-logicism - Formal logic and reasoning
• review-taylorism - Workflow and efficiency analysis

Fixer MCP Tools

Three fixer strategies are available:

• fix-comment - Add TODO comments marking issues for later fixing
• fix-conservative - Fix only clear-cut, low-risk issues that won't require cascading changes
• fix-zealot - Pick one important issue and fix it comprehensively, even if it requires major reorganization

Development

Testing

• npm test - Run all test suites
• npm run test:mcp - Run MCP protocol validation
• npm run test:comprehensive - Run comprehensive feature tests

Utilities

• npm run status - Display server capabilities and status
• npm run check - Check VS Code integration readiness

Manual Testing

To debug, you can run the server manually:

• npm start - Run the server in standalone mode
• npm run dev - Run in development mode with auto-restart

Project Structure

mcp-review/
├── src/           # Server implementation
├── tests/         # Test suites
├── scripts/       # Utility scripts
├── ai-review/     # Critic and fixer frameworks (submodule)
└── README.md      # Project documentation

Credit

Critic frameworks provided by the ai-review repository by Joe Nelson.