Prime MCP Server

Model Context Protocol (MCP) server for Prime Intellect - manage GPU pods, check availability, and control compute resources through MCP.

Features

• GPU Availability Checking - Search for available GPU instances across providers
• Pod Management - Create, list, monitor, and delete GPU pods
• Cluster Support - Check multi-node cluster availability
• SSH Key Management - Manage SSH keys for pod access

Demo

Installation

pip install prime-mcp-server

Or with uv:

uv pip install prime-mcp-server

Configuration

Prime MCP uses prime-core for configuration, which supports multiple authentication methods:

Option 1: Environment Variable

export PRIME_API_KEY="your-api-key"

Option 2: Prime CLI Login (Recommended)

prime login

This stores your API key in ~/.prime/config.json and is shared across all Prime tools.

Option 3: Config File

Manually edit ~/.prime/config.json:

{
  "api_key": "your-api-key",
  "base_url": "https://api.primeintellect.ai"
}

Quick Start

Running the MCP Server

python -m prime_mcp.mcp

The server runs over stdio and can be integrated with MCP clients (like Claude Desktop, IDEs, or other MCP-compatible tools).

Available Tools

For complete documentation of all available tools and their parameters, see docs.primeintellect.ai.

Development

# Clone the repo
git clone https://github.com/PrimeIntellect-ai/prime.git
cd prime/packages/prime-mcp-server

# Install dependencies
uv sync

# Run tests
uv run pytest

# Run linter
uv run ruff check .

Related Packages

• prime - Full CLI + SDK with pods, sandboxes, inference, and more
• prime-sandboxes - SDK for managing remote code execution environments
• prime-evals - SDK for managing model evaluations