Prerequisites
- Python: Universal MCP requires Python 3.8 or higher. You can check your Python version by running
python --version. - pip: Ensure you have pip, the Python package installer, available. It usually comes with Python.
- uv (Optional but Recommended for Application Packages): For installing application packages dynamically (used by
app_from_slug), Universal MCP utilizesuv. Ifuvis not found in your PATH, the system will attempt to use common installation paths. Ensureuvis installed and accessible for the best experience with dynamic application loading. You can installuvby following the instructions on its official website.
Installing Universal MCP
You can install the core Universal MCP package using pip:Installing Specific Application Integrations
Universal MCP offers a convenient CLI command to help set up configurations for specific applications that you might want your AI agents to interact with, especially when using the AgentR platform. This often involves setting up how the MCP server connects to these applications. The command is:<app_name> with the name of the application you wish to install (e.g., claude, cursor, github). This command typically configures your local environment to point to the Universal MCP system, often involving an AgentR API key.
Refer to cli.py for more information.
For a list of currently supported applications for this installation method, you can often find them by running the command without an app name or with a non-existent one, or refer to the utils/installation.py file in the SDK.
Supported applications include: claude, cursor, cline, continue, goose, windsurf, zed.
Note: While universal_mcp install helps with client-side configuration for certain pre-defined apps, the core way to make tools available to your MCP server (especially a LocalServer) is by configuring them in your server’s configuration file, which we’ll cover next. Dynamically loaded applications via app_from_slug also rely on uv for package installation.
With Universal MCP installed, you’re ready to configure and run your first server!