🔌 MCP Server Manager

The Model Context Protocol (MCP) is becoming the standard for connecting AI agents to external tools. Managing a growing collection of MCP servers manually is friction. This removes it.

The Problem

As MCP adoption grows, agent developers accumulate more servers: filesystem, GitHub, databases, custom tools. There’s no lightweight tool to register them, check their health, or export configurations across projects and tools.

The Solution

MCP Server Manager is a CLI + Python library for managing MCP server registries. Register servers, check health, filter by tag, and import/export Claude Desktop configurations with a single command.

Features:

• 3 transport types: stdio, http, sse
• Health checking: binary detection for stdio, HTTP GET for http/sse
• Claude Desktop integration: import existing configs, export for new installs
• Tag-based filtering: group servers by project, environment, or capability
• JSON output: pipe to jq or other tools

Usage

# Register servers
mcp-manager add filesystem --transport stdio --command npx -a @modelcontextprotocol/server-filesystem /data
mcp-manager add github --transport http --url http://localhost:3000 --tag dev
mcp-manager add my-db --transport stdio --command python -a server.py --tag prod --tag db

# Inspect
mcp-manager list
mcp-manager list --tag prod
mcp-manager health

# Claude Desktop integration
mcp-manager import-claude    # Import from Claude Desktop config
mcp-manager export-claude    # Export for Claude Desktop

# Show details
mcp-manager show github      # Full config
mcp-manager show github --json  # Machine-readable

from mcp_manager import MCPRegistry, MCPServer, Transport

registry = MCPRegistry()
server = MCPServer(
    name="my-tool",
    transport=Transport.STDIO,
    command="python",
    args=["server.py"],
    tags=["prod"],
)
registry.add(server)

healthy = [s for s in registry.list() if registry.health(s.name)]

Status

✅ Complete — 37/37 tests passing

No external dependencies. Python 3.9+. MIT License.