I Built a Package Manager for MCP Servers in One Day

Ever tried setting up MCP servers for Claude Code or Cursor? You end up copy-pasting JSON configs, hunting for package names, and praying the args are right.

I got tired of it and built mcpx — a pip-installable CLI that manages MCP servers the way npm manages packages.

3-Line Install

pip install mcpx
mcpx search filesystem
mcpx install filesystem

That's it. The server gets added to your ~/.claude.json (or Cursor/VS Code config) automatically. Restart your AI tool and it works.

What It Does

• 50+ curated MCP servers across 21 categories (databases, cloud, dev tools, AI, etc.)
• Auto-detects your platform — Claude Code, Cursor, VS Code, or project-level .mcp.json
• Smart search — find servers by name, description, tags, or category
• One-command install/uninstall — handles config JSON, env vars, and parameters
• Doctor command — diagnoses config issues, missing deps, broken servers
• Beautiful CLI — built with typer + rich for a polished terminal experience

Architecture

mcpx/
├── cli.py          # typer CLI with 11 commands
├── config.py       # Auto-detect Claude/Cursor/VS Code configs
├── registry.py     # 50+ server registry with search/browse
├── installer.py    # Install/uninstall into JSON configs
├── doctor.py       # Diagnostic checks (deps, config, platforms)
└── registry_data.json  # Curated server database

The registry is a curated JSON file with server metadata — package name, command, args, env vars, config parameters, categories, and star counts. The installer resolves parameters (like API keys) into the final config entry.

Example Workflow

# See what is popular
mcpx top

# Search for database servers
mcpx search database

# Get details
mcpx info supabase

# Install with API key
mcpx install supabase -p api_key=your_key

# Check your setup
mcpx doctor

# List what is installed
mcpx list

Why I Built This

MCP (Model Context Protocol) is Anthropic's open standard that lets AI tools like Claude connect to external services — databases, APIs, file systems, cloud providers. It is powerful but the setup is manual: find the right npm package, figure out the args, edit JSON config, hope it works.

I wanted the experience of brew install or npm install but for MCP servers. Search, install, done.

The Tech Stack

• Python 3.9+ with full type hints
• typer for the CLI framework
• rich for beautiful terminal output
• pytest with 84% coverage (46 tests)
• GitHub Actions CI across Python 3.9-3.13
• ruff + black for linting and formatting

Try It

pip install mcpx
mcpx top

GitHub: github.com/LakshmiSravyaVedantham/mcpx

Stars appreciated if you find it useful!

Inspired by the MCP ecosystem growth and projects like claude-code-templates. Built to make the MCP developer experience as smooth as possible.