mcp-ephemeral-k8s

What is mcp-ephemeral-k8s?

A Python library for spawning ephemeral Model Context Protocol (MCP) servers on Kubernetes using Server-Sent Events (SSE). It supports multiple runtimes (Node.js via npx, Python via uvx) and works with mcp-proxy.

How to use mcp-ephemeral-k8s?

Run the server with uvx mcp-ephemeral-k8s or install as a Python package (pip install mcp-ephemeral-k8s and then mcp-ephemeral-k8s). Optionally deploy via a Helm chart. Connect using an SSE URL (e.g., http://localhost:8000/sse) with the sse transport.

Key features of mcp-ephemeral-k8s

• Supports Node.js (via npx) and Python (via uvx) runtimes
• Integrates with mcp-proxy for runtime execution
• Works with local kubeconfig or in-cluster configuration
• Can be run as an MCP server itself
• Provides a Python library for programmatic session management
• Exposes ephemeral MCP servers over SSE

Use cases of mcp-ephemeral-k8s

• Spawning temporary MCP servers on Kubernetes for development or testing
• Running MCP servers with different runtimes in isolated ephemeral pods
• Automating MCP server lifecycle with a Kubernetes-based scheduler
• Integrating MCP-based tools into CI/CD pipelines on Kubernetes

FAQ from mcp-ephemeral-k8s

What transport does the server use?

SSE (Server-Sent Events), as shown in the connection URL http://localhost:8000/sse and configuration example with "transport": "sse".

Does it require a Kubernetes cluster?

Yes, it requires a Kubernetes cluster with valid kubectl configuration. A local cluster (e.g., Kind) can be used for development.

How can I connect to an ephemeral MCP server?

After spawning, the server provides an SSE URL. Configure your MCP client with that URL and transport set to sse.

What runtimes are supported?

Node.js (via npx) and Python (via uvx). The library works with mcp-proxy to run these runtimes.

What are the prerequisites?

Docker and a Kubernetes cluster (e.g., Kind or any cluster with kubectl). For installation via uvx or pip, Python is also required.