Recognize when a task exceeds Runpod's current management-API surface and answer honestly — state the gap plainly, name the specific operation that isn't exposed, and give the real working path. Never fabricate a tool call or claim a silent success. Trigger on phrases like "store a secret / gated Llama token", "set up…
Host a ComfyUI workflow as a serverless endpoint and prove it with one generated image. Owns the workflow-as-job-input pattern (the workflow JSON rides each job request, it is never baked into the image), worker-variant selection by the models the workflow references, and custom-model provisioning (Civitai…
Read-only Runpod money hygiene — find idle spend, explain serverless idle-worker billing, break spend down by resource, and give a whole-account snapshot. Trigger on phrases like "why is my runpod bill high", "what am I paying for", "idle pods", "serverless charges with no traffic", "cost breakdown", "spend by…
Answer "what's available, what fits, and what does it cost" against Runpod's catalogs — GPU types with price, live stock and per-data-center availability, the capacity matrix across host CUDA versions, the Hub worker marketplace, and the pay-per-use public-endpoints catalog. Read-only, never mutates; answers against…
Operate a Serverless endpoint that already exists — diagnose stuck jobs and dying workers, tune its scaling and cold-start settings safely, and manage its queue. Read-only by default; mutates only the exact setting the user authorized. Trigger on phrases like "my jobs are stuck INQUEUE", "the video job never runs"…
Create, verify (read-back), and manage the lifecycle of Runpod's reusable plumbing resources — templates, network volumes, and container registry credentials — plus caching model weights on a volume and diagnosing a private-registry pull failure. Trigger on phrases like "create a template", "reusable template", "make…
Provision a Runpod Pod for interactive, dev, or training use and tell the user exactly how to connect to it. Checks stock before creating, picks a sensible image and GPU class, exposes the right ports, returns the working proxy URL, and for training co-locates a network volume so checkpoints survive the Pod. Also…
Diagnose a Runpod Pod that is unreachable, GPU-blind, or stuck, then perform a recovery action only if the user authorizes one. Trigger on phrases like "pod 502", "can't reach my pod", "torch can't see the GPU", "cuda not available", "pod came back with no gpu", "did my model load", "read my pod logs", "is anything…
Start here for any Runpod task — running GPU/CPU pods, deploying serverless endpoints, templates, network volumes, catalog and cost questions. Routes the request to the right Runpod journey skill and defines the answer contract every Runpod reply follows. All work happens through the connected Runpod MCP tools.
Bring any source — a HuggingFace repo, a ComfyUI workflow, a Hub release, a custom container image, a GGUF repo, a Civitai asset, or a community template — to a proven Runpod Serverless endpoint. Deploys an existing source through the Runpod MCP tools; code-first Python deploys (the runpod-flash dev loop) are outside…
★not rated 79▲
+1 todayA190 tokens
At most 3 mods per repository are shown here, and a mod shipped inside a plugin is left to that plugin's page — the rest are on their repository pages: