Diagnostics MCP Server

HTTP MCP server with 5 diagnostic tools providing real-time access to ALL VS Code diagnostics (TypeScript, ESLint, Prettier, and all installed extensions)

🎯 Overview

This Model Context Protocol (MCP) server provides AI agents with real-time access to all diagnostics from your VS Code workspace, including:

• ✅ TypeScript/JavaScript errors and warnings
• ✅ ESLint linting issues
• ✅ Prettier formatting issues
• ✅ All Language Servers (Python, Go, Rust, etc.)
• ✅ All VS Code Extensions diagnostics
• ✅ Real-time updates as you code

⚙️ Configuration

The extension provides two configuration settings:

diagnostics-mcp-server.autoStart

• Type: boolean
• Default: true
• Description: Automatically start HTTP MCP server when VS Code opens

To disable auto-start:

1. Open VS Code Settings (Ctrl+,)
2. Search for "diagnostics-mcp-server"
3. Uncheck "Auto Start"
4. Use the "Start HTTP MCP Server" command to start manually

diagnostics-mcp-server.port

• Type: number
• Default: 3846
• Description: Port for HTTP MCP server

To change the port:

1. Open VS Code Settings (Ctrl+,)
2. Search for "diagnostics-mcp-server.port"
3. Set your desired port number
4. Restart the server or reload VS Code
5. Update your MCP client configuration with the new port

📋 Installation

Step 1: Install VS Code Extension

Install from VS Code Marketplace:

1. Open VS Code
2. Go to Extensions (Ctrl+Shift+X)
3. Search for "Diagnostics MCP Server"
4. Click Install

Latest Version: 1.0.12 - Complete HTTP MCP implementation with 5 diagnostic tools, enhanced error handling, and working commands

Step 2: Extension Auto-Start

The extension automatically starts the HTTP MCP server when VS Code opens. No additional setup required!

Server Details:

• Protocol: HTTP with Server-Sent Events
• Port: 3846 (automatically managed)
• Startup: Automatic with VS Code

Step 3: Configure MCP Client

Add this to your MCP client configuration (e.g., Claude Desktop config or VS Code MCP settings):

{
  "mcpServers": {
    "diagnostics": {
      "type": "http",
      "url": "http://127.0.0.1:3846/mcp",
      "description": "VS Code diagnostics - all 5 tools (errors, warnings, info, health, all diagnostics)"
    }
  }
}

Verify Connection

1. Check server status: Visit http://127.0.0.1:3846/health
2. View logs: VS Code Output panel → "Diagnostics MCP Server"
3. Test connection: Server automatically starts when VS Code opens

Available immediately after VS Code extension installation - no additional setup required!

Usage

Once configured, AI agents (like Claude, GitHub Copilot) can use these 5 MCP tools:

1. get_all_diagnostics - Get complete diagnostic information from workspace
2. get_errors - Get only error-level diagnostics
3. get_warnings - Get only warning-level diagnostics
4. get_info - Get only info-level diagnostics
5. get_workspace_health - Get workspace health score (0-100)

🔧 How It Works

This extension uses an HTTP MCP Server architecture:

┌─────────────────────────────────────────────────────┐
│  AI Agent (Claude, GitHub Copilot)                 │
│  ↓                                                  │
│  HTTP MCP Protocol (port 3846)                     │
│  ↓                                                  │
│  VS Code Extension (HTTP MCP Server)               │
│  ↓                                                  │
│  vscode.languages.getDiagnostics() API             │
│  ↓                                                  │
│  ALL Diagnostics (TS, ESLint, Prettier, etc.)     │
└─────────────────────────────────────────────────────┘

Why Extension Required?

• VS Code diagnostics are only accessible inside VS Code via the vscode module
• The extension provides the bridge between VS Code APIs and the MCP server
• This ensures you get ALL diagnostics from ALL sources, not just TypeScript

📦 What's Included

• HTTP MCP Server - Runs on port 3846 with Server-Sent Events
• 5 Diagnostic Tools - Comprehensive workspace diagnostic access
• 3 VS Code Commands - Start/Stop/Status server control
• Real-time Updates - Live diagnostic monitoring
• Health Scoring - Workspace quality metrics (0-100)

🛠️ Development

Build from Source

git clone https://github.com/Maaz0313-png/Diagnostics-MCP.git
cd "Diagnostics MCP"
npm install
npm run compile

Test Locally

# Test the launcher
node index.js --help

# Test with a workspace
node index.js

📖 API Reference - 5 MCP Tools

1. Tool: get_all_diagnostics

Get complete diagnostic information from workspace.

Returns:

{
  "total": 42,
  "diagnostics": [
    {
      "file": "src/app.ts",
      "line": 10,
      "column": 5,
      "severity": "error",
      "message": "Type 'string' is not assignable to type 'number'",
      "source": "ts"
    }
  ],
  "status": "found",
  "timestamp": "2025-10-02T10:30:00.000Z"
}

2. Tool: get_errors

Get only error-level diagnostics.

Returns:

{
  "count": 5,
  "diagnostics": [...],
  "severityLevel": "errors",
  "status": "found",
  "timestamp": "2025-10-02T10:30:00.000Z"
}

3. Tool: get_warnings

Get only warning-level diagnostics.

Returns:

{
  "count": 3,
  "diagnostics": [...],
  "severityLevel": "warnings",
  "status": "found",
  "timestamp": "2025-10-02T10:30:00.000Z"
}

4. Tool: get_info

Get only info-level diagnostics.

Returns:

{
  "count": 2,
  "diagnostics": [...],
  "severityLevel": "info",
  "status": "found",
  "timestamp": "2025-10-02T10:30:00.000Z"
}

5. Tool: get_workspace_health

Get workspace health score (0-100) based on diagnostics.

Returns:

{
  "healthScore": 85,
  "status": "good",
  "summary": {
    "errors": 2,
    "warnings": 5,
    "infos": 3,
    "total": 10
  },
  "timestamp": "2025-10-02T10:30:00.000Z"
}

Health Score Calculation:

• Errors: -10 points each
• Warnings: -3 points each
• Info: -1 point each
• Scale: 0-100 (100 = perfect health)
• Status: excellent (90+), good (70+), fair (50+), poor (<50)

🎮 VS Code Commands

Four commands available in Command Palette (Ctrl+Shift+P):

1. 🚀 Diagnostics MCP: Start HTTP MCP Server

   • Manually start the MCP server
   • Use if server didn't auto-start or autoStart is disabled

2. 🛑 Diagnostics MCP: Stop HTTP MCP Server

   • Stop the running MCP server
   • Useful for troubleshooting

3. 🔄 Diagnostics MCP: Restart HTTP MCP Server

   • Restart the MCP server (stop + start)
   • Use after changing configuration settings (port, etc.)

4. 📊 Diagnostics MCP: MCP Server Status (5 Tools + Health)

   • View server status, current diagnostics count, and health score
   • Quick health check of your workspace

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

📄 License

MIT License - see LICENSE file for details

🔗 Links

• GitHub Repository
• Model Context Protocol

⚠️ Troubleshooting

"MCP server not connecting"

1. Check server status: Visit http://127.0.0.1:3846/health
2. View logs: VS Code Output panel → "Diagnostics MCP Server"
3. Restart server: Use command "Diagnostics MCP: Start HTTP MCP Server"
4. Reload VS Code window: Ctrl+Shift+P → "Reload Window"

"Port 3846 already in use"

1. Stop other applications using port 3846
2. Or change port in VS Code settings: diagnostics-mcp-server.port
3. Use "Restart HTTP MCP Server" command or reload VS Code
4. Update your MCP client config with the new port

"No diagnostics returned"

1. Open a workspace with code files
2. Wait for language servers to initialize
3. Check VS Code's Problems tab for diagnostics

📝 Version History

1.0.14 (Current)

• ✅ Configuration settings support (autoStart, port)
• ✅ Restart command for easy server restart
• ✅ Configurable port number
• ✅ Optional auto-start disable

1.0.12-1.0.13

• ✅ Complete HTTP MCP server implementation
• ✅ 5 specialized diagnostic tools
• ✅ Enhanced error handling and connection stability
• ✅ Working VS Code commands (Start/Stop/Status)
• ✅ Comprehensive tool documentation in metadata
• ✅ Beautiful diagnostic icon
• ✅ Full workspace health scoring

1.0.11

• ✅ Enhanced connection stability for empty diagnostics
• ✅ HTTP transport implementation

1.0.10

• ✅ Added severity-specific tools (get_errors, get_warnings, get_info)

1.0.0 (Initial Release)

• ✅ Basic VS Code diagnostics integration
• ✅ Support for all language servers and extensions

💡 Use Cases

• AI-Powered Code Review: Let AI agents analyze all code issues
• Automated Quality Checks: Monitor workspace health in real-time
• Smart Refactoring: AI can see all diagnostics before suggesting changes
• Learning Assistant: Help users understand and fix code issues
• CI/CD Integration: Pre-commit diagnostic analysis

---

Made with ❤️ by Maaz Tajammul