Rent a Cloud GPU with Hermes Agent or OpenClaw

Description and use

A CPU‑based bot rents a GPU instance on Vast.ai via Hermes Agent or OpenClaw, runs the job over SSH, downloads results, and destroys the instance, keeping costs low.

Problem

Bots running on CPU servers need GPU compute for specific jobs but lack GPU resources locally.

Outcome

A single chat message can trigger a full GPU rental cycle: the bot quotes, rents, runs, downloads, and cleans up an instance in under 4 minutes, costing only $0.007 for the test render.

Workflow

• Describe job and limits to bot
• Bot searches Vast.ai offers and returns a quote
• User approves quote
• Bot generates SSH key, registers it with Vast, creates instance with specified Docker image
• Bot waits for SSH readiness
• Bot uploads inputs, verifies uploads via SHA‑256
• Bot confirms GPU usage with nvidia‑smi and Blender Cycles CUDA check
• Bot runs job over SSH
• Bot downloads outputs, verifies integrity
• Bot destroys instance, deletes SSH key
• Bot confirms instance removal and key deletion

Evidence

The whole rental, from creating the machine to releasing it, took 194 seconds. Vast posted a $0.007 charge for that instance.

Source

The bot then compared the GPU frame with its earlier CPU preview and told us honestly what had improved and what still fell short, rather than declaring the job done.

Source

The bot uploads its inputs, installs or unpacks the software, and runs the job. Two checks separate a real GPU run from an expensive CPU run: Verify the uploads. Comparing a hash of each file on both ends catches a truncated upload before you pay for a broken run.

Source

The job is only finished when this job's instance ID no longer appears in your instance list and its temporary SSH key is gone.

Source