CoreHu

CoreHu is a pay-per-second GPU cloud for AI work. This connector lets Claude search live GPU offers (RTX 4090 to A100, H100, H200 and B200), see how many machines are free per GPU model and country, compare starting prices and estimate the cost of a job. Describe what you want to run — serve Llama, Qwen or Mistral with vLLM or Ollama, fine-tune a model with LoRA, generate images with ComfyUI, or train with PyTorch — and Claude estimates the GPU memory you need and suggests the cheapest machines that fit, with a ready setup recipe.

After you sign in, you choose on a consent screen what Claude may access: selected machines (status, GPU/CPU metrics, usage, SSH host and port), optionally your USD balance and, only if you allow it, renting, stopping and starting machines. You can set optional limits (hourly price, machines at once, rentals per day). Every rental is prepared first with the exact hourly price and only happens after you approve it in the chat; Claude then follows the startup and gives you the SSH command, port tunnels or JupyterLab link. Claude cannot add funds, make payments or delete machines. Access expires after 30 days and can be revoked anytime in the CoreHu console under Keys → AI connections.

Try: "Recommend a GPU to serve Llama 3 70B in 4-bit", "Is there a free H100 in the US?", "What would 10 hours on an RTX 4090 cost?", "Rent it", "Is my machine ready?", "Stop my machine".