RunPod MCP Server

RunPod MCP Server is an MCP server that connects Claude and other Model Context Protocol clients to the RunPod REST API for managing cloud compute infrastructure. Machine learning engineers, developers, and DevOps teams use this integration to automate and control GPU workloads without having to switch back and forth to the RunPod web console. The server exposes dedicated tools to create, inspect, update, start, stop, and terminate standard GPU and CPU pods directly within natural language conversations. It also supports configuration and lifecycle management for serverless endpoints, customizable environment templates, persistent network storage volumes, and container registry authentication records. By linking your RunPod account via an API key, the server translates natural language requests into direct cloud provisioning actions, streamlining remote GPU instance management, batch compute tasks, and model inference deployments straight from your local MCP client interface.

Category: AI & LLM Tooling

Tags: cloud, compute, gpu, infrastructure, runpod

Visit RunPod MCP Server

How to install and configure RunPod MCP Server

1. Ensure Node.js 18 or higher is installed on your system. 2. Clone the repository and navigate into the directory: bash git clone https://github.com/antonioevans/runpod-mcp-ts cd runpod-mcp-ts 3. Install project dependencies and build the server: bash npm install npm run build 4. Open your Claude for Desktop configuration file located at ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows). 5. Add the server entry under the mcpServers object, supplying your RunPod API key: json { "mcpServers": { "runpod": { "command": "node", "args": ["/path/to/runpod-mcp-server/build/index.js"], "env": { "RUNPOD_API_KEY": "your_api_key_here" } } } } 6. Save the file and restart Claude for Desktop. Alternatively, install automatically via Smithery with npx -y @smithery/cli install @runpod/runpod-mcp-ts --client claude.

What you can do with RunPod MCP Server

• Provisioning on-demand GPU instances with specified container images, GPU types, and counts directly through conversational prompts. * Inspecting, starting, stopping, and terminating running RunPod pods to monitor cloud workloads and optimize compute expenses. * Deploying and adjusting serverless inference endpoints by setting template identifiers along with minimum and maximum worker limits. * Creating and managing persistent network volumes to retain model weights, training datasets, and outputs across pod restarts. * Managing container registry authentications to allow secure image pulls from private registries during pod setup.

Key facts

• https://github.com/antonioevans/runpod-mcp-ts
• AI & LLM Tooling, Cloud & Infrastructure
• cloud, compute, gpu, infrastructure, runpod

Part of MCP Servers

What is RunPod MCP Server?

RunPod MCP Server is a Model Context Protocol integration that exposes tools for managing RunPod cloud GPU resources, serverless endpoints, network volumes, and templates through an MCP-compatible client like Claude for Desktop.

How do I install RunPod MCP Server?

You can install it automatically using Smithery by running the npx installation command targeting Claude Desktop, or manually by cloning the GitHub repository, running npm install and npm run build, and adding the built index.js path with your RUNPOD_API_KEY environment variable to your Claude Desktop configuration file.

What prerequisites are needed to run RunPod MCP Server?

Running this server requires Node.js version 18 or higher, an active RunPod account with an API key generated from the RunPod console, and an MCP client such as Claude for Desktop.

Is RunPod MCP Server open source?

Yes, RunPod MCP Server is open source and distributed under the MIT license on GitHub.

Which MCP clients work with RunPod MCP Server?

The server works with Claude for Desktop and any other MCP-compliant client that supports standard Model Context Protocol server configurations over local command execution.