This MCP server enables comprehensive management of RunPod cloud infrastructure through the Model Context Protocol, allowing AI assistants like Claude to programmatically interact with the RunPod REST API.

Capabilities:

• Pod Management: Create, list, retrieve details, update, start, stop, and delete pods with configurable options including GPU type/count, container images, Docker settings, environment variables, ports, storage volumes, data centers, and container disk sizes. Filter pods by name, GPU type, compute type, and data center.

• Serverless Endpoint Management: Create, list, retrieve details, update, and delete serverless endpoints with auto-scaling configurations (min/max workers, scaler type, idle timeout), GPU configuration, data center selection, and template-based deployment.

• Template Management: Create, list, retrieve details, update, and delete templates for reusable container configurations with Docker settings, environment variables, volumes, ports, and serverless options.

• Network Volume Management: Create, list, retrieve details, update (name and size), and delete network volumes (1-4000 GB) for persistent storage across data centers.

• Container Registry Authentication: Create, list, retrieve details, and delete container registry authentication credentials (username and password) for accessing private Docker images.

Required runtime environment for the MCP server implementation

Enables management of NVIDIA GPU-powered computing resources through RunPod's platform

Supports deployment of PyTorch environments through RunPod's container infrastructure

Facilitates creation and management of Ubuntu-based development environments on RunPod