runpod-mcp: Instructions file for Claude Code

CLAUDE.md

runpod-mcp CLAUDE.md is an instructions file for Claude Code from kyle-nelson-berkeley/runpod-mcp. It costs 3,139 tokens per session, scanned A, original, MIT.

A Claude Code instruction file for a RunPod subfolder containing a FastMCP server that lets agents operate a GPU training runbook. RunPod is a cloud service for renting computing machines with GPUs.

In plain words
What is it for?
It is for setting up and controlling RunPod pods, launching and monitoring training jobs, applying BlueROV2 patches, syncing logs, checking GPU availability, and reporting spend.
Why use it?
It documents the folder's tools, operating rules, and safeguards for managing training jobs and GPU resources.

Instructions file for Claude Code

Written for Claude Code: the file is CLAUDE.md. Also seen: mentions CLAUDE.md.

This is kyle-nelson-berkeley/runpod-mcp's own configuration. It tells Claude Code how to work on runpod-mcp itself, so it is not a mod to install elsewhere. Copy it as a starting point and replace the rules that are about this project. Everything runpod-mcp configures →

Reuse

Borrowing it

Nothing to install: this file belongs to kyle-nelson-berkeley/runpod-mcp. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.

Copy the file
curl -O https://raw.githubusercontent.com/kyle-nelson-berkeley/runpod-mcp/main/CLAUDE.md
Clone the repo
git clone --depth 1 https://github.com/kyle-nelson-berkeley/runpod-mcp

Made for: Claude Code.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for runpod-mcp CLAUDE.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/kyle-nelson-berkeley/runpod-mcp/claude-md.svg)](https://agentmods.dev/instructions/kyle-nelson-berkeley/runpod-mcp/claude-md)
Your own site
<a href="https://agentmods.dev/instructions/kyle-nelson-berkeley/runpod-mcp/claude-md"><img src="https://agentmods.dev/badge/instructions/kyle-nelson-berkeley/runpod-mcp/claude-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 3,139 This file is loaded in full into every session.
When invoked 3,139 The same file — it is already loaded in full.
Security scan A 0 findings. A grade says what 26 rules found in the file — not that it is safe.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.03139 $0.03139
Opus 5 $0.01570 $0.01570
Sonnet 5 $0.00628 $0.00628
Haiku 4.5 $0.00314 $0.00314

Measured 7d ago against content hash 2e4b9f5a075b, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-07, from the pricing page.

Security

Grade A, and why

runpod-mcp CLAUDE.md scanned grade A with 0 findings against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 7d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Nothing flagged

None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.

CLAUDE.md · 190 lines

How it starts

The opening of the file, as written. The whole thing — 190 lines — stays where its author put it; the contents beside it link to each section on GitHub.

runpod-mcp — CLAUDE.md

Last updated: 2026-08-04 · Owner: Kyle

Project context, operating rules, and the runbook-day → tool map live in the root CLAUDE.md — read that first. This file covers only this subfolder.

A · What this folder is

A self-contained FastMCP server (14 tools) that lets agents drive the entire Learning-to-Swim GPU runbook on RunPod: pod lifecycle (ensure_pod/stop_pod/terminate_pod), setup (run_pod_setup), detached training jobs (launch_training, run_job, job_status), BlueROV2 patching (apply_bluerov_patches, axis_sanity_sweep), plus sync_logs / spend_report / pod_status / gpu_availability / exec_on_pod. Cost guardrails are enforced in code, not by convention: one pod PER VEHICLE (unknown pod names refused), RTX 4090 only, Secure Cloud, never interruptible, network volume required, idle watchdog, per-job auto_stop.

Watchdog status (2026-08-09): the pod-side idle watchdog has never once armed in recorded history — the install probe failed on every attempt (opaque rc=91 pre-fix). Mechanism facts: runpodctl is IMAGE-SHIPPED (a container-disk wipe restores the disk from the image, it does not empty it — the binary is back on every boot); what the wipe removes is runtime-installed material (idle_watchdog.sh itself, apt libs, rsync), which is why install-on-every-transition stays. The 2026-08-09 fix ships a three-way diagnostic probe (NO_RUNPODCTL exit 90 / NO_RUNPODCTL_AUTH_BARE diagnostic-continue / NO_RUNPODCTL_AUTH_SOURCED exit 91 / PROBE_OK) with /etc/rp_environment sourced unconditionally; a success reports armed (stop path unverified) — the probe certifies READ, the watchdog needs WRITE; first real confirmation is a successful stop in /workspace/.idle_watchdog.log. Defect 2 ships DIAGNOSED, not CLOSED — the next bring-up's sentinel settles H1 (no credential on the pod) vs H2 (unsourced shell, fixed by the sourcing change). Until then: idle_watchdog: FAILED ⇒ arm the Mac-side deadman before any job.

Read the full file on GitHub · 190 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 7d ago First seen · 190 lines · 3,139 tokens per session scan A 2e4b9f5a075b

Subscribe to this mod's changes

runpod-mcp CLAUDE.md is an instructions file published in the GitHub repository kyle-nelson-berkeley/runpod-mcp (0 stars, last pushed 10d ago), licensed MIT. It adds 3,139 tokens to every session, about $0.0157 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-31.

Related

Other instructions, from other repositories