Configuration

Configuration Guide

Configure kit-dev-mcp for your development environment

Environment Variables

LLM API Keys (Required for get_code_summary)

An LLM API key is required for the get_code_summary tool. Choose one of:

# Option 1: OpenAI (default) export OPENAI_API_KEY="sk-..."
# Option 2: Anthropic export ANTHROPIC_API_KEY="sk-ant-..."
# Option 3: Google Gemini export GOOGLE_API_KEY="AI..."

Note: All other tools work without an LLM API key.

KIT_GITHUB_TOKEN

GitHub personal access token for accessing private repositories and increasing API rate limits.

export KIT_GITHUB_TOKEN="ghp_your_token_here"
KIT_CACHE_DIR

Directory for caching symbol extraction and documentation. Defaults to ~/.cache/kit

export KIT_CACHE_DIR="~/.cache/kit"
CHROMA_PACKAGE_SEARCH_API_KEY

API key for Chroma Package Search - enables searching through source code of popular packages. Get your key from Chroma Cloud.

export CHROMA_PACKAGE_SEARCH_API_KEY="your_api_key"

Enables: package_search_grep, package_search_hybrid, package_search_read_file

MCP Server Configuration

Cursor Configuration

Add to your Cursor settings.json

{
  "mcpServers": {
    "kit-dev": {
      "command": "uvx",
      "args": ["--from", "cased-kit[all]", "kit-dev-mcp"],
      "env": {
        "OPENAI_API_KEY": "sk-...",              // For get_code_summary (or use ANTHROPIC_API_KEY)
        "KIT_GITHUB_TOKEN": "ghp_..."            // Optional: for private repos
      }
    }
  }
}

Advanced Configuration Options

Environment Variables for Advanced Users

{
  "env": {
    "KIT_GITHUB_TOKEN": "ghp_...",           // GitHub access token
    "KIT_CACHE_DIR": "~/.cache/kit",         // Cache directory  
    "KIT_LOG_LEVEL": "DEBUG",                // Logging level
    "KIT_MAX_WORKERS": "8",                  // Parallel workers
    "KIT_CACHE_TTL": "3600",                 // Cache TTL in seconds
    "KIT_DEEP_RESEARCH": "true"              // Deep doc research
  }
}

KIT_LOG_LEVEL: Set to DEBUG for verbose logging

KIT_MAX_WORKERS: Number of parallel workers (default: 4)

KIT_CACHE_TTL: Cache time-to-live in seconds

KIT_DEEP_RESEARCH: Enable deep documentation research

File Watching Configuration

Configure which files to watch and ignore:

# .kitignore file in your project root
node_modules/
.git/
*.pyc
__pycache__/
.venv/
dist/
build/
*.log
.DS_Store

Performance Tuning

Cache Settings
  • • Enable aggressive caching for large repos
  • • Set appropriate TTL values
  • • Use local SSD for cache directory
  • • Clear cache periodically
Memory Management
  • • Adjust worker count based on CPU cores
  • • Limit file size for analysis
  • • Configure max memory usage
  • • Enable incremental processing