Overview
This MCP server enables programmatic interaction with Hyperbolic's GPU cloud from a local or cloud environment. It exposes GPU discovery, rental, and SSH tooling through a self-contained Node.js server, facilitating agents and LLMs to manage GPU resources and workloads. Prerequisites include Node.js 16+, npm or yarn, a Hyperbolic API token, and optionally an SSH private key for connecting to GPU instances. The setup guides users through creating a Hyperbolic account, funding their balance, generating an API token, and adding an SSH public key. After installation (clone, npm install, npm run build), start the server locally with npm start. The server can integrate with Claude for Desktop using a small JSON config that points to the built entry script and provides environment variables (HYPERBOLIC_API_TOKEN and SSH_PRIVATE_KEY_PATH). Available commands cover GPU management (list-available-gpus, rent-gpu-instance, terminate-gpu-instance, list-user-instances, get-cluster-details) and SSH tooling (ssh-connect, remote-shell, ssh-disconnect). Environment variables can be supplied via Claude config or a .env file. Actions incur charges on the Hyperbolic account, and the SSH private key must be unencrypted.
Features
list-available-gpus
Lists all available GPUs on the Hyperbolic network.
rent-gpu-instance
Rents a GPU instance from a specific cluster; requires cluster_name, node_name, and gpu_count.
terminate-gpu-instance
Terminates a GPU instance by its instance_id.
list-user-instances
Lists all active GPU instances rented by the user.
get-cluster-details
Fetches detailed information about a specific cluster using cluster_name.
ssh-connect
Establishes an SSH connection to a remote GPU instance using host, username, and authentication details.
remote-shell
Executes a command on the connected remote GPU server.
ssh-disconnect
Closes the active SSH connection.
Who Is This For?
- Developers:Programmatically manage GPU availability, rent GPUs, and SSH into instances.
- ML Engineers:Orchestrate GPU-backed experiments and training jobs via the MCP server.
- AI Agents:Interact with GPUs through the MCP to view, rent, and run tasks.
