Runpod
Connect Runpod so agents can check and manage your Pods and Serverless endpoints.

The Runpod MCP server lets agents look into a stuck job or a failed run without you opening the console.
For use cases and example prompts, see Runpod for AI coding agents.
What is Runpod?
Runpod is a cloud platform for running GPU and CPU workloads, on Pods or on Serverless endpoints.
What agents can do
Once you connect Runpod, agents can:
- List, create, update, delete, start, and stop Pods, and read their logs
- List, create, update, and delete Serverless endpoints, and check their health and workers
- Submit, check, cancel, and retry jobs, and clear an endpoint's queue
- Create and manage clusters, templates, and network volumes
- Manage secrets, container registry credentials, and SSH keys
- Read billing history
- Search the Runpod docs
Agents can only do what the connected API key and its owner's Runpod role allow.
With a key that allows it, agents can delete Pods and endpoints, clear queued jobs, change credentials, and start work that costs money. Use a Read Only key if agents should only look.
Runpod's server also has optional tools that send Runpod feedback, questions, saved lessons, and a note on how you found Runpod. It tells agents to send that note once per account without asking you, and to skip these tools for sensitive work.
Setup
On the API Keys tab of the Runpod console's Credentials page, click Create API Key, choose its permissions, and copy the key. See Runpod's API key docs for each permission. Your ability to create an API key depends on your role in your Runpod organization.
Then connect it in Superconductor: find Runpod on the Connectors page, click Connect, and paste your API key. The Connectors guide covers the sharing and project access options on that form.
Reconnecting and disconnecting
If you rotate or delete your key, agents' Runpod calls fail until you save a new one. Click Manage next to the connection, enter the new key, and click Save.
To disconnect, click Manage, then Disconnect, then delete the key on the API Keys tab of the Runpod console.
When to use it
Ask agents to use Runpod for a stuck job, a failing endpoint, or a Pod you need to check on. For example:
Check the status of this Runpod job and tell me why it failed.Read our training Pod's logs and tell me why the last run failed.Summarize our Runpod billing by day for last month.Search the Runpod docs for how network volumes work, then list the ones we have.