MCP-manager
A centralized MCP (Model Context Protocol) server management and routing system that allows you to manage multiple MCP servers through a single interface with fine-grained access control.
claude mcp add --transport stdio mykolavaskevych-mcp-manager python main.py mcp-manager.yaml \ --env NPM_TOKEN="optional-npm-auth-token" \ --env DATABASE_URL="your-database-connection-string" \ --env GITHUB_TOKEN="optional-github-token" \ --env WEATHER_API_KEY="your-weather-api-key"
How to use
MCP Manager provides a centralized control plane for multiple MCP servers. It can automatically discover and install backend MCP servers from sources like NPM, GitHub, or local paths, route requests between clients and servers, and enforce per-client access rules. Use the manager to start, monitor, and configure all managed MCP servers from a single YAML file. Clients (such as VSCode MCP Extension or Claude Desktop) connect to the manager rather than to individual MCP servers, simplifying authentication and permission management. The manager exposes tooling to start, stop, install, validate configurations, and inspect permissions, making it easier to maintain secure and scalable MCP environments.
How to install
Prerequisites:
- Python 3.8+ and pip
- Access to install pip dependencies
Install the MCP Manager from source:
- Clone the repository or download the source code.
- Install dependencies in editable mode (as recommended in the README):
cd MCP-manager
pip install -e .
Alternative (if you prefer to run via CLI directly):
cd MCP-manager
uv sync
Create and customize your configuration:
cp mcp-manager.example.yaml mcp-manager.yaml
# Edit mcp-manager.yaml to configure your servers, clients, sources, and runtime options
Run the MCP Manager:
python main.py mcp-manager.yaml
# or using the CLI (if available in your environment)
mcp-manager start --config mcp-manager.yaml
Verify installation and configuration:
mcp-manager config validate
mcp-manager config show
Additional notes
Notes and tips:
- This MCP Manager is described as archived in the README; prefer running via Docker if available to simplify deployments and isolation.
- The configuration supports automatic installation of backend MCP servers from NPM, GitHub, or local paths. Ensure your sources are reachable and tokens (NPM/GitHub) are set if required.
- Environment variables such as WEATHER_API_KEY and DATABASE_URL may be needed by certain servers; keep them in sync with your actual backend requirements.
- If you encounter access issues, review client permissions with the provided CLI commands (e.g., mcp-manager client permissions <client_id>). Use config validate to catch syntax or schema problems early.
- For production, consider running behind a reverse proxy, enabling health checks, and tuning cache and concurrency settings in runtime configuration.
Related MCP Servers
MCP-Bridge
A middleware to provide an openAI compatible endpoint that can call MCP tools
mcp-google-ads
An MCP tool that connects Google Ads with Claude AI/Cursor and others, allowing you to analyze your advertising data through natural language conversations. This integration gives you access to campaign information, performance metrics, keyword analytics, and ad management—all through simple chat with Claude, Cursor or Windsurf.
mcp-in-action
极客时间MCP新课已经上线!超2000同学一起开启MCP学习之旅!
mcp-rquest
A MCP server providing realistic browser-like HTTP request capabilities with accurate TLS/JA3/JA4 fingerprints for bypassing anti-bot measures. It also supports converting PDF and HTML documents to Markdown for easier processing by LLMs.
asterisk
Asterisk Model Context Protocol (MCP) server.
mcpd
A ~3MB MCP daemon that exposes any script as a tool for Claude, Cursor & AI assistants