Skip to main content

Getting Started

This guide walks through installing MCP Compose, creating a configuration, and launching the server with Web UI and REST API.

Prerequisites​

  • Python 3.10+
  • Node.js 18+ (only required if you build the Web UI from source)
  • Docker (optional)

Install​

From PyPI​

pip install mcp-compose

From source​

git clone https://github.com/datalayer/mcp-compose.git
cd mcp-compose
pip install -e .
git clone https://github.com/datalayer/mcp-compose.git
cd mcp-compose
docker-compose up -d

Verify installation​

mcp-compose --version

Quick Start​

Create a file named mcp_compose.toml with a minimal configuration:

[composer]
name = "my-composer"
conflict_resolution = "prefix"

[transport]
stdio_enabled = true

[[servers.proxied.stdio]]
name = "filesystem"
command = ["uvx", "mcp-server-filesystem", "/data"]
working_dir = "${MCP_COMPOSE_CONFIG_DIR}"

[[servers.proxied.stdio]]
name = "calculator"
command = ["uvx", "mcp-server-calculator"]
working_dir = "${MCP_COMPOSE_CONFIG_DIR}"

Start MCP Compose:

mcp-compose serve --config mcp_compose.toml

Once running, you have access to:

CLI Quick Actions​

List all available tools across your composed servers:

mcp-compose list-tools --config mcp_compose.toml

Invoke a tool directly from the command line:

mcp-compose invoke-tool filesystem_read_file '{"path": "/data/readme.txt"}' --config mcp_compose.toml

Next Steps​

  • Read the Architecture guide to understand how MCP Compose works
  • Explore the Configuration reference for all available options
  • See the Clients guide to connect Claude Desktop or VS Code
  • Configure authentication for production deployments