# ParalonCloud Rentals [Health: Active]

**Category:** 💻 Developer Tools  
**Repository:** https://github.com/ParalonCloud/mcp-rentals  
**GitHub Stars:** 2  
**Views:** 0  
**Installs:** 0  
**Upvotes:** 0  
**Directory Page:** https://allmcps.com/mcp/paraloncloud-rentals

## Description
Rent GPUs from ParalonCloud: browse, start, poll, and stop GPU rentals over the rental API.

## Tools
Capabilities this server exposes over MCP:

- **list_gpus** — List rentable GPUs with price, VRAM, compute capability, country
- **get_balance** — Your credit balance
- **create_rental** — Start a Jupyter rental (async → poll)
- **get_rental** — Status + connection URL once running
- **list_rentals** — Your active rentals (`status: "all"` for history)
- **destroy_rental** — Stop a rental and stop billing

## Claude Desktop Quick Installation
Install path detected from listing signals. Uses `npx` (confidence: high):

```json
"mcpServers": {
  "paraloncloud-rentals": {
    "command": "npx",
    "args": ["-y","@paraloncloud/mcp-rentals"],
    "env": {
      "PARALON_API_KEY": ""
    }
  }
}
```

**Requires environment variables:** `PARALON_API_KEY` — the values above are empty placeholders; fill in real credentials before running (see the repository for what each one is for).

## Documentation

## What ParalonCloud Rentals MCP server does

ParalonCloud Rentals MCP server exposes the ParalonCloud Rental API through MCP tools. An agent can inspect available GPU nodes, check the account balance, create a Jupyter rental, wait for it to become ready, retrieve its connection URL, and stop it when work is complete.

The server is intended for temporary GPU workloads managed from an AI client. A typical interaction starts by comparing available GPUs, choosing one based on price or hardware requirements, and creating a rental for a defined number of hours. The agent can then use the returned Jupyter connection details and later terminate the rental.

## How it works

The MCP client starts the server with the ParalonCloud API key in `PARALON_API_KEY`. Tool calls are sent to the ParalonCloud Rental API. `create_rental` is asynchronous, so the agent uses `get_rental` to check progress and obtain the connection URL after the rental is running.

The available tools are:

- `list_gpus`: Returns rentable GPUs with price, VRAM, compute capability, and country.
- `get_balance`: Reads the account's credit balance.
- `create_rental`: Starts a Jupyter rental and spends credits.
- `get_rental`: Reports rental status and provides the connection URL when ready.
- `list_rentals`: Lists active rentals, or rental history when called with `status: "all"`.
- `destroy_rental`: Stops a rental and ends its billing.

The server automatically creates an idempotency key for rental creation, preventing a retry from starting a second GPU. Supplying `hours` gives the rental an automatic stop time.

## Setup and configuration

Create an API key in the ParalonCloud Console and enable the GPU Rentals scope. The README recommends using a dedicated agent key rather than a production key. The Console also provides a `max_active_rentals` limit that caps concurrent rentals.

Configure the MCP client to run `npx -y @paraloncloud/mcp-rentals` and pass the key through `PARALON_API_KEY`. Claude Desktop and Cursor use an MCP server configuration with this command and environment variable; Claude Code can add the server with `claude mcp add`.

`PARALON_BASE_URL` is optional and defaults to `https://paraloncloud.com/api/v1`. For local development, the server can be started with `node server.js` while exporting the API key.

## Limitations and notes

Creating and destroying rentals change what the account is billed. MCP clients should keep tool-call confirmation enabled; Claude Code and Claude Desktop confirm these calls by default. A key's rental scope must be enabled, and the account must have sufficient credits for creation.

ParalonCloud Rentals MCP server manages Jupyter rentals through the rental API; the provided material does not describe tools for provisioning other environments. Rental duration, automatic stopping, and the account's active-rental cap should be considered when designing an agent workflow.

_Full upstream README: https://allmcps.com/mcp/paraloncloud-rentals/readme_