Runpod MCP server

What the model can do

Runpod GPU cloud: pods with GPU, image, status, hourly cost and live utilisation, serverless endpoints with workers, scaling and request metrics, network volumes, GPU types with prices and stock, and spend by day, pod or endpoint; when you ask, can create, start, stop, restart, change or terminate a pod, rescale or delete a serverless endpoint, and create, grow or delete a network volume.

Reads

• runpod_list_podsPods in Runpod, newest first: name, status (RUNNING, EXITED, ERROR…), GPU type and count or CPU flavour, image, hourly cost in USD, cloud…
• runpod_get_podOne Runpod pod by id, with ports, disks, mounted network volumes, CUDA version and, while it runs, live GPU, CPU and memory utilisation and uptime.
• runpod_list_endpointsServerless endpoints: name, queue or load-balancer type, image, GPU pools and GPUs per worker (or CPU), minimum and maximum workers, idle timeout…
• runpod_endpoint_workersThe workers of one serverless endpoint right now: running, idle, initialising, throttled or unhealthy, with GPU type, data centre…
• runpod_endpoint_metricsRequest metrics for one serverless endpoint over a window (default the last 24 hours, at most 90 days back): requests, completed…
• runpod_spendRunpod spend in USD over a window (default the last 30 days), in day, week, month…
• runpod_gpu_typesGPU types Runpod offers, with VRAM, hourly list price per GPU on secure and community cloud and serverless, the most per pod, and…
• runpod_network_volumesNetwork volumes: name, size in GB, data centre and storage tier.

Acts, when you ask

These change something in Runpod. The model is told to use them only when you ask for exactly that, and each one is granted by name: per team on a shared connection, by the owner on a personal one.

• runpod_create_podCreates a Runpod pod on one GPU type (or CPU flavour) and starts it: hourly billing at that GPU's rate begins as soon as it is placed…
• runpod_start_podStarts a stopped (EXITED or ERROR) Runpod pod, which resumes GPU billing at its hourly rate.
• runpod_stop_podStops a running or provisioning Runpod pod: compute is released and stops being billed, its disk is kept (and still billed)…
• runpod_restart_podRestarts a running Runpod pod's container in place, interrupting whatever it is running; the ephemeral container disk is wiped.
• runpod_terminate_podTerminates and deletes a Runpod pod.
• runpod_update_podChanges a Runpod pod: its name, image, container disk size, exposed ports, or locks it (a locked pod cannot be stopped or reset).
• runpod_update_endpointChanges a Runpod serverless endpoint: minimum and maximum workers, idle timeout, request timeout, FlashBoot, image or name.
• runpod_delete_endpointDeletes a Runpod serverless endpoint and its bound template.
• runpod_create_network_volumeCreates a Runpod network volume in one data centre; its size is billed monthly from creation whether or not anything mounts it…
• runpod_update_network_volumeRenames a Runpod network volume or grows it (size can only increase, and the larger size is billed from then on).
• runpod_delete_network_volumeDeletes a Runpod network volume and every file on it.

Connect it

First connect Runpod on Dashboard → Connections, then add Memtro to your client once. The Runpod tools appear alongside the tools of everything else you have connected; nothing is configured per connector.

claude plugin marketplace add memtro-org/claude-memtro
claude plugin install memtro@memtro --config api_key=sk-memtro-...

{
  "mcpServers": {
    "memtro": {
      "url": "https://platform.memtro.com/api/mcp",
      "headers": {
        "Authorization": "Bearer sk-memtro-..."
      }
    }
  }
}

{
  "mcp": {
    "memtro": {
      "type": "remote",
      "url": "https://platform.memtro.com/api/mcp",
      "headers": {
        "Authorization": "Bearer sk-memtro-..."
      }
    }
  }
}

Who may use it

The key identifies the person, and the server offers that person the tools they may use everywhere else in Memtro. A shared Runpod connection is used by the teams an administrator granted it to, tool by tool, on the Permissions page; a personal connection is used by its owner alone, and its tools that act stay off until the owner allows them. See Teams and permissions and Connectors.

Ask it like this

• › Which pods are running and what do they cost per hour?
• › What did we spend on Runpod last month?
• › How busy is the GPU on the training pod?
• › How many requests failed on the inference endpoint today?
• › What does an H100 cost per hour on secure cloud?

Questions

Is this an official Runpod MCP server?

No. It is Memtro's MCP server with Memtro's Runpod connector behind it: the connector reads Runpod through its API with the credentials you connect on Dashboard → Connections, and Runpod publishes nothing here. The same endpoint carries every other system you connect and Memtro's memory tools.

What does it cost?

Nothing on top of Memtro. One price per organisation by its number of users, from $50 a month for up to 10 users after a 7-day free trial, with every connector, the MCP server, jobs and memory included. You bring your own model keys.

Can it change anything in Runpod?

Only when you ask, and only what it has been granted: create a pod (starts hourly GPU or CPU billing); start a stopped pod (resumes billing); stop a running pod (keeps its disk); restart a running pod's container; terminate a pod (permanent); rename a pod, change its image, disk, ports or lock; rescale or change a serverless endpoint; delete a serverless endpoint (permanent); create a network volume (billed storage); rename or grow a network volume; delete a network volume (permanent). Every other tool reads, searches or lists. An administrator grants each acting tool by name per team, and a personal connection acts only after its owner allows it.

Which clients work?

Any MCP client that speaks Streamable HTTP: Claude Code, Cursor, Claude Desktop, OpenCode, Codex, Zed, Windsurf, Cline and the Anthropic and OpenAI SDKs. There is no local process to run; the server is remote and authenticated with your key.

Other MCP servers · Engineering and operations

• GitHub MCP server
• AWS CloudWatch MCP server
• Sentry MCP server
• ServiceNow MCP server
• Okta MCP server
• Twilio MCP server