magg

Magg is a meta MCP server that acts as a central hub for discovering, configuring, and proxying multiple MCP servers. It provides a unified interface to search for new MCP tools, install and configure them, and expose their capabilities under consolidated prefixes. With Magg, you can dynamically add or disable MCP servers, automatically aggregate their tools, and persist configuration across sessions. It also includes built-in commands like magg_status and magg_check to monitor health and readiness, making it easier to manage a growing toolkit of LLM capabilities.

To use Magg, install it as a Python tool and run the CLI. By default, Magg runs in stdio mode (for integration with desktop LLM clients). You can also serve over HTTP to enable system-wide access. Once running, Magg scans configured MCP servers, proxies their tools, and exposes them under a unified namespace. The Magg CLI supports handy commands and transports, including real-time MCP notifications, making it straightforward to add new servers and keep configurations in sync without restarting the service.

Prerequisites:

• Python 3.12 or higher (3.13+ recommended)
• uv (recommended) for running Magg as a tool (optional but recommended)

Install Magg as a Python package and run the CLI:

# Prerequisite: ensure Python 3.12+ is installed
python3 --version

# Option 1: Install Magg as a tool via uv (recommended way per docs)
# Install uv if you don't have it yet (on macOS/Linux):
# curl -sSf https://install.astral.sh/uv | sh

uv tool install magg

# Run Magg in stdio mode (default)
magg serve

# Run Magg with HTTP transport (exposes an HTTP API)
magg serve --http

Alternative: Run Magg directly from GitHub without installation:

# Run with stdio transport
uvx --from git+https://github.com/sitbon/magg.git magg

# Run with HTTP transport
uvx --from git+https://github.com/sitbon/magg.git magg serve --http

Local development (clone and install in editable mode):

git clone https://github.com/sitbon/magg.git
cd magg

# Install in development mode with dev dependencies
uv sync --dev
# or with poetry
poetry install --with dev

# Run the CLI to verify
magg --help

Docker (optional):

# Run production image (default port 8000)
docker run -p 8000:8000 ghcr.io/sitbon/magg:latest

Tips and common considerations:

• Magg stores configuration in .magg/config.json by default; you can customize the location via MAGG_CONFIG_DIR (see environment variables).
• If you enable HTTP transport, consider enabling MAGG_PRIVATE_KEY for RSA-based JWT authentication to secure HTTP access.
• Use the built-in magg_status and magg_check tools to monitor health and performance.
• Docker users can mount a custom config directory into /home/magg/.magg or use the MAGG_CONFIG_DIR environment variable to point Magg to a specific directory.
• When aggregating tools from multiple MCP servers, Magg proxies them under unified prefixes to simplify access for clients that don’t support dynamic updates.
• Ensure Python 3.12+ is used to take full advantage of Magg’s features and compatibility.