code-quality-mcp
# š Code Quality MCP Server
> **Hybrid Code Quality Analysis** - Combines logic-based checks with AI-powered deep analysis
MCP (Model Context Protocol) system for code quality analysis. Supports two modes:
- ā” **Fast Mode**: Quick logic-based checks
- š¤ **Deep Mode**: AI-powered deep analysis (Claude)
---
## š§ Installation
```bash
npm install
npm run build
```
## ā¶ļø Running the Server
### Option 1: Direct Run
```bash
npm start
```
### Option 2: Via Claude Desktop
Add to `~/.config/Claude/claude_desktop_config.json`:
```json
{
"mcpServers": {
"code-quality": {
"command": "node",
"args": ["/absolute/path/to/code-quality-mcp/dist/index.js"]
}
}
}
```
---
## š API Tools
### š Analysis Tools
#### 1. `check_quality` - Full Code Quality Check
```typescript
{
"tool": "check_quality",
"arguments": {
"projectPath": ".",
"deepAnalysis": false, // true = AI mode
"checkUnusedCode": true,
"checkComplexity": false, // requires AI
"checkSecurity": false, // requires AI
"page": 1, // pagination
"pageSize": 50 // items per page
}
}
```
**Returns:** Complete list of issues with pagination
---
### š” Smart Tools (Recommended!)
#### 2. `get_smart_summary` - Smart Summary ā
Instead of 500 issues, get a focused summary:
```typescript
{
"tool": "get_smart_summary",
"arguments": {
"projectPath": "."
}
}
```
**Returns:**
```
š Score: 68/100
š Issues: 500 (120 critical)
ā±ļø Fix Time: 2 days
š„ Top Problems:
⢠unused-code: 300 (60%)
⢠security: 50 (10%)
š Hotspot Files:
š“ UserService.ts - 45 issues
```
---
#### 3. `get_quick_wins` - Quick Wins ā”
Get only actions that provide maximum impact in minimum time:
```typescript
{
"tool": "get_quick_wins",
"arguments": {
"projectPath": "."
}
}
```
**Returns:**
```
ā” Quick Wins:
1. Remove 50 unused vars (10 min) ā +15 points
2. Fix 30 console.log (5 min) ā +8 points
3. Translate Hebrew comments (15 min) ā +5 points
Total: 30 minutes ā +28 points!
```
---
#### 4. `get_trends` - Progress Tracking š
See how code quality improves over time:
```typescript
{
"tool": "get_trends",
"arguments": {
"projectPath": "."
}
}
```
**Returns:**
```
š Trends:
ā
Score: 45 ā 68 (+23)
ā
Fixed: 120 issues
ā ļø New: 15 issues
Improving:
⢠unused-code: 300 ā 180 (-120)
⢠security: 10 ā 5 (-5)
Degrading:
⢠code-style: 50 ā 65 (+15)
```
---
### š§ Other Tools
#### 5. `analyze_project` - Project Type Detection
```typescript
{
"tool": "analyze_project",
"arguments": {
"projectPath": ".",
"deep": true
}
}
```
#### 6. `get_recommendations` - Quick Recommendations
```typescript
{
"tool": "get_recommendations",
"arguments": {
"projectPath": ".",
"language": "en" // or "he" for Hebrew
}
}
```
---
## ā” Fast vs Deep Mode
| Feature | Fast Mode | Deep Mode |
|---------|-----------|-----------|
| Speed | ~20ms | ~1000ms |
| Cost | Free | API calls |
| Accuracy | Good | Excellent |
| AI insights | ā | ā
|
**When to use Fast Mode:**
- During development
- Before commits
- In CI/CD pipelines
**When to use Deep Mode:**
- Before Pull Requests
- Code reviews
- Security audits
---
## š Examples
- **[HOW_TO_USE.md](./HOW_TO_USE.md)** - šÆ How to use this MCP in Claude (commands, examples)
- **[examples/usage-examples.ts](./examples/usage-examples.ts)** - Complete API examples
---
## š Supported Platforms
- Node.js / TypeScript
- React / React Native
- Next.js / NestJS
- Firebase / AWS Amplify
- Java / .NET / Angular
---
## š Key Features
ā
**Smart Ignore Patterns** - Automatically skips:
- `node_modules`, `build`, `dist`, `.git`
- React Native: `android/gradle`, `ios/Pods`, native build folders
- Only scans **your code**, not auto-generated files
ā
**Pagination** - Handle large projects with many issues
ā
**Smart Summaries** - Get actionable insights instead of overwhelming lists
ā
**Quick Wins** - Find high-impact, low-effort fixes
ā
**Trend Tracking** - Monitor code quality improvements over time
---
**Built with ā¤ļø using MCP and Claude**
TDQS
Scored across 6 tools
analyze_project and check_quality overlap in purpose, both analyzing project type and quality. The get_recommendations, get_smart_summary, and get_quick_wins tools are related but distinct in output style, so disambiguation is moderate.
All tool names consistently follow a verb_noun pattern (analyze_project, check_quality, get_recommendations, etc.). The use of get_ for most retrieval functions is uniform and predictable.
Six tools is within the ideal 3-15 range for a focused code-quality server. Each tool addresses a specific aspect of the domain without bloat.
The tool set covers analysis, quality checking, recommendations, summaries, quick wins, and trends, providing a comprehensive view of code quality. Missing a raw issue list or fix application is a minor gap, but agents can work around it via the existing tools.