Give your AI agent a GPU in one command

Coding agents already run shell commands, so the lium CLI is all they need. The agent rents a pod, runs code, copies results back and removes the pod. No browser, no human click. Per-second billing means a short GPU test costs cents.
pip install lium.io && lium up --gpu H100 --ttl 1h
In our test with Claude Code, an H100 pod was online in under 10 seconds after lium up, and the run cost a few cents.

Set up in three steps

  1. On a machine the agent cannot read (e.g. your laptop when the agent runs on a server), create a budgeted key: pip install lium.io, lium workspaces login, then lium keys create agent --scope read --scope rent --pod-visibility own --daily-budget 20. Never log in where the agent runs: the login stores your own session in ~/.lium. It needs an email and password; signed up with Google or GitHub? Set a password in Settings.
  2. On the agent's machine, install the CLI and set only that key: export LIUM_API_KEY=sk_..., then check it with lium whoami. The key sees only its own pods, stops at its budget and has no billing scope.
  3. Add the rules below to CLAUDE.md (Claude Code) or AGENTS.md (Codex), then ask, e.g. "benchmark this training script on a GPU and give me the numbers".
## GPU compute (Lium)

Rent GPUs with the `lium` CLI. The API key is configured; never print it.
- Pick a node with `lium ls --count 1 --format json` (cheapest first) unless I ask for a specific GPU.
- Always rent with a name and a deadline: `lium up <huid> --name <name> --ttl 1h -y --json`.
- Run work with `lium exec <name> "<command>"`; use `lium exec <name> -d "<command>"` for anything long.
- Bring results back with `lium scp <name> <remote-file> <local-dir> -d`.
- When done, or if anything fails: `lium rm <name> -y`, then `lium ps` to confirm.
- If a rent is refused with API_KEY_BUDGET_EXCEEDED, stop and tell me.

Guardrails the server enforces

  • Scoped keys. A key has only the scopes it was created with. Card payments and credit transfers need billing; no key can create keys or raise its own budget.
  • Budgets per key. Daily, monthly or total. A key at its budget cannot rent (it gets 402 API_KEY_BUDGET_EXCEEDED), and its pods are removed when spend reaches it. Set it at creation or later with lium keys budget (API keys).
  • Its own pods only. With --pod-visibility own a key sees and controls only the pods it rented.
  • Automatic removal. --ttl schedules removal at creation, so a crashed or forgotten session leaves no pod running (scheduled termination).
  • A real machine. Root access and its own Docker daemon, so the agent can build and run containers.

Tools built for agents

  • Agent skill: npx skills add Datura-ai/lium-skill --skill lium.
  • Docs MCP server: https://docs.lium.io/mcp.
  • Full agent reference in one fetch: lium.io/llms-full.txt.
  • Headless signup, funding and the REST API: AI agents docs.
For AI agents (CLI install, agent skill, headless signup): llms-full.txt