Portable MCP Toolkit

# Local-First MCP Servers Setup ## Status ### ✅ Already Configured 1. **filesystem** - `@modelcontextprotocol/server-filesystem` - Status: ✅ Installed and configured - Provides: Direct file I/O operations - Location: `.cursor/mcp.json` (lines 13-20) ### 🔍 Available but Need Manual Setup The following servers are not in the official MCP Registry but can be installed manually: 2. **pydantic-ai/mcp-run-python** - Python sandbox - Package: Likely `pydantic-ai` Python package - Installation: `pip install pydantic-ai` - Note: May need to check GitHub for MCP server implementation 3. **Qdrant/mcp-server-qdrant** - Vector memory for RAG - Alternative found: `@rocklerson/openai-mcp-qdrant` - Official: Check Qdrant GitHub for official MCP server - Note: Your code-intelligence server already uses Qdrant locally 4. **datalayer/jupyter-mcp-server** - Jupyter notebook automation - Found alternative: `io.github.ChengJiale150/jupyter-mcp-server` (not in registry) - Installation: May need to install from GitHub ## Current Configuration Your `.cursor/mcp.json` includes: ```json { "mcpServers": { "code-intelligence": { "command": "cmd", "args": ["/c", "python", "C:\\AIStack-MCP\\mcp_intelligence_server.py", "--workspace", "C:\\AIStack-MCP"] }, "filesystem": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-filesystem", "C:\\AIStack-MCP"] }, "github": { "command": "npx", "args": ["-y", "@modelcontextprotocol/server-github"], "env": { "GITHUB_PERSONAL_ACCESS_TOKEN": "${GITHUB_PERSONAL_ACCESS_TOKEN}" } } } } ``` ## Recommendations ### Option 1: Use What You Have Your current setup already provides: - ✅ **File I/O**: `filesystem` server - ✅ **Vector Search**: `code-intelligence` server (uses Qdrant) - ✅ **Code Analysis**: `code-intelligence` server (uses Ollama) ### Option 2: Add Missing Servers Manually If you need the specific servers mentioned: 1. **Python Sandbox (pydantic-ai)**: ```powershell pip install pydantic-ai # Then check GitHub for MCP server wrapper ``` 2. **Qdrant MCP Server**: ```powershell npm install -g @rocklerson/openai-mcp-qdrant # Or check Qdrant GitHub for official server ``` 3. **Jupyter Server**: ```powershell # Check GitHub repositories for installation # Example: https://github.com/ChengJiale150/jupyter-mcp-server ``` ## Next Steps 1. **Test current setup** - Your filesystem and code-intelligence servers are working 2. **Identify specific needs** - Determine which additional servers you actually need 3. **Install from source** - If servers aren't in registry, install from GitHub 4. **Add to config** - Update `.cursor/mcp.json` with new server configurations ## Notes - The MCP Registry may not have all community servers - Some servers are only available via npm/pip directly - GitHub repositories often have installation instructions - Your code-intelligence server already provides vector search (Qdrant) and code analysis (Ollama)