MCP Servers Overview

MCP (Model Context Protocol) servers expose tools and resources that AI models can use during inference, deployed as Kubernetes Deployments and accessible via a Streamable HTTP Transport endpoint.

What is MCP?

The Model Context Protocol is an open standard for connecting AI models to external tools, data sources, and services. An MCP server implements a JSON-RPC 2.0 API that a client (like the Admin Chat assistant) can call to:

• Execute code
• Query databases
• Call external APIs
• Access file systems
• Run computations

MCP Server List

The MCP Servers page shows all deployed instances with:

Column	Description
Name	Server name (also the URL slug)
Status	Pod status: Running, Starting, Pending, Failed
Endpoint	URL of the MCP HTTP endpoint
Usage	Real-time CPU (millicores) and memory (MB)
Actions	Open endpoint, delete

Built-in Admin MCP Server

Menatic AI includes a built-in Admin MCP Server used by the Admin Chat assistant. It provides tools for managing the cluster:

• list_jupyter_notebooks — list deployed notebooks
• create_jupyter_notebook — deploy a new notebook
• list_mcp_servers — list deployed MCP servers
• create_mcp_server — deploy a new MCP server
• get_dashboard_stats — get cluster statistics
• list_ai_models — list available models

This server runs as a separate HTTP handler within the backend process and is not listed on the MCP Servers page.

User-Deployed MCP Servers

Users can deploy custom MCP servers from the MCP Servers page. Any Docker image that implements the MCP Streamable HTTP Transport protocol can be deployed.

See Deploying MCP Servers for instructions.