Runpod
Let your coding agent start, resize, and tear down RunPod GPU pods or serverless endpoints without leaving the chat.
Overview
com.mcparmory/runpod is a MCP server for the Operate phase that lets agents launch, scale, and manage RunPod GPU pods and serverless endpoints across regions.
What is this MCP server?
- Launch and manage dedicated GPU pods across RunPod regions from MCP tool calls
- Configure and scale serverless GPU endpoints for inference or batch jobs
- Ship via stdio using PyPI package mcparmory-runpod (uvx) or OCI image ghcr.io/mcparmory/runpod:1.0.1
- Registry version 1.0.1 with dual transport paths (uvx + Docker) for local or containerized agents
- Repository sourced from github.com/mcparmory/registry for MCP Armory catalog consistency
- Server schema version 1.0.1
- 2 distribution packages: PyPI mcparmory-runpod (uvx) and OCI ghcr.io/mcparmory/runpod:1.0.1
- Transport type: stdio
Community signal: 25 GitHub stars.
Who is it for?
Indie builders already on RunPod who want Claude Code or Cursor to handle pod and serverless lifecycle during ship-and-operate loops.
Skip if: Teams that only train on local GPUs, use a different cloud exclusively, or are not ready to store RunPod API keys for stdio MCP.
What do I get? / Deliverables
Your agent can create, scale, and retire RunPod resources in-thread so GPU spend and endpoint config stay visible in the same session as your code changes.
- Agent-callable control of RunPod pods and serverless endpoints
- Repeatable infra changes documented in the agent session transcript
- Aligned registry package at version 1.0.1 (PyPI + OCI)
Recommended MCP Servers
Journey fit
RunPod control is ongoing production infrastructure work—scaling GPUs and endpoints after you ship—not one-off product coding. Infra is the canonical shelf for cloud GPU orchestration, region selection, and endpoint lifecycle outside your laptop.
How it compares
RunPod MCP integration for live infra control, not a training notebook skill or a generic Dockerfile generator.
Common Questions / FAQ
Who is com.mcparmory/runpod for?
Solo and small-team builders on RunPod who ship agent-assisted apps and need programmatic control of GPU pods and serverless endpoints from their IDE agent.
When should I use com.mcparmory/runpod?
Use it during Operate (and late Build) when you are provisioning inference GPUs, switching regions, or scaling serverless endpoints instead of clicking through the RunPod UI.
How do I add com.mcparmory/runpod to my agent?
Register the MCP server in Claude Code, Cursor, or another stdio-capable client using uvx with identifier mcparmory-runpod from PyPI, or the Docker image ghcr.io/mcparmory/runpod:1.0.1, with your RunPod API key in environment or client secrets.