An agent with shell access can rent a GPU on Lium without a browser. It reads the docs, lists machines as JSON, launches a pod with a spend cap and a time limit, runs its job, and removes the pod. Lium’s documentation calls the CLI “the preferred surface for an agent: it covers signup, funding, renting and SSH, so there is no web form to fill and no HTTP to hand-roll.” This guide walks through that workflow and the three features that make it safe to automate. The commands come from documentation and a Lium team article, not from an independent test, so check them against your installed CLI.
What the agent can read: llms.txt and friends
Lium’s AI Agents documentation lists several resources meant for machines:
As an Amazon Associate I earn from qualifying purchases.
- llms.txt and llms-full.txt: plain-text documentation indexes for language models (for example,
lium.io/llms.txt). - A docs MCP endpoint: lets an agent search the documentation for a specific question instead of loading everything.
- An OpenAPI spec: for REST operations the CLI does not cover.
- A public pricing JSON feed and a public nodes feed: for current prices and availability.
lium ls --format json: the machine list as JSON from the CLI.
The docs treat the CLI as the first choice and the REST API as a fallback for anything the CLI lacks.
The workflow, step by step
1. Sign up and keep the recovery credential
Fingerprint signup needs no browser. It returns a 32-character recovery credential once, and the docs say to store it immediately because there is no password reset. The API key can come back null if minting fails, so have the agent check for that. Email signup or an existing account may involve different steps. The docs say new-account signup credit depends on abuse checks and configuration, so don’t plan around receiving one.
#1 Best Overall
- Axial-tech fans now feature a smaller fan hub that facilitates longer blades and a barrier ring that increases downward air pressure
- 2.5-slot design allows for greater build compatibility while maintaining cooling performance
- 0dB technology lets you enjoy light gaming in relative silence
- Dual BIOS switch lets you toggle between Quiet and Performance BIOS profiles
- Dual ball fan bearings last up to twice as long as sleeve bearing designs
2. Fund the account
The balance must exceed 15 minutes of the chosen node’s hourly price before you can rent. At $2.00 an hour, for example, that is a balance above $0.50. (This is an illustrative calculation, not a quoted price.)
3. List machines as JSON
Have the agent run lium ls --format json and pick a node by ID. Don’t hard-code prices. Prices and availability change, so read them from the live feed or the listing at the moment of the decision.
Rank #2
- Powered by the NVIDIA Blackwell architecture and DLSS 4
- Powered by GeForce RTX 5070 Ti
- Integrated with 16GB GDDR7 256bit memory interface
- PCIe 5.0
- WINDFORCE cooling system
4. Launch with guard rails
The official docs outline a non-interactive launch: pass the node ID to lium up with --yes --no-ssh and --ttl, so the pod removes itself. Then run work with lium exec.
Free tools Windows power users keep installed
One-click scans. No signup required.
The Lium team’s article shows a second form, written for Lium CLI 0.8.0:
Rank #3
- Powered by the NVIDIA Blackwell architecture and DLSS 4
- Powered by GeForce RTX 5060
- Integrated with 8GB GDDR7 128bit memory interface
- PCIe 5.0
- WINDFORCE cooling system
lium up --gpu H100 -y --json --budget 12.50 --ttl 3h --timeout 6005. Run the task
Use lium exec to run commands on the pod, or SSH if the agent needs a session. The agent should capture output and write results somewhere that outlives the pod.
6. Remove the pod
Run lium rm <pod>. The team article says removal stops the meter, and it describes per-second billing until then. Treat that as the article’s description and confirm it in the current billing docs before depending on it.
Rank #4
- Powered by Radeon RX 9070 XT
- WINDFORCE Cooling System
- Hawk Fan
- Server-grade Thermal Conductive Gel
- RGB Lighting
What each flag does
| Flag | Purpose (per the Lium team article, CLI 0.8.0) |
|---|---|
--gpu H100 |
Selects a matching H100 node |
-y / --yes |
Skips the interactive confirmation |
--json |
Emits machine-readable output the agent can parse |
--budget 12.50 |
Terminates the pod at the specified spend |
--ttl 3h |
Terminates after the elapsed time |
--timeout 600 |
Bounds how long the command waits for the rental to complete |
Budget and TTL are separate limits. Set both, and whichever is reached first ends the rental. Run lium up --help first, since flags and units can change between versions. The article’s example was based on version 0.8.0 and lium.io/llms.txt as read on September 24, 2026.
Example prices, with a warning
The Lium team article listed these starting prices per GPU-hour, observed at 02:44 UTC on September 24, 2026:
Best Value
- Axial-tech fans now feature a smaller fan hub that facilitates longer blades and a barrier ring that increases downward air pressure
- Phase-change GPU thermal pad helps ensure optimal heat transfer, lowering GPU temperatures for enhanced performance and reliability
- 2.5-slot design allows for greater build compatibility while maintaining cooling performance
- Dual-ball fan bearings last up to twice as long as standard conventional sleeve bearings designs
- 0dB technology lets you enjoy light gaming in relative silence
| GPU | From (per GPU-hour) |
|---|---|
| RTX 4090 | $0.45 |
| RTX 5090 | $0.58 |
| H100 80GB HBM3 | $1.30 |
| H200 | $3.00 |
| B200 | $5.60 |
| B300 | $8.25 |
These are a company-published snapshot, not current offers. Use the live pricing feed for today’s numbers.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.Safety checklist for unattended agents
- Always set both
--budgetand--ttl; never launch without a cap. - Store the recovery credential and API key outside the agent’s chat transcript.
- Make the agent verify the pod is gone (for example by listing pods) after
lium rm. - Save outputs off the pod before removal.
- Have the agent read the installed CLI’s help instead of trusting remembered flags.
- Choose a budget you would accept losing if the task hangs.
Comparing with other GPU rental services
Only Lium’s side is documented here, so there is no ranking. If you compare providers, check the same six things for each: which GPUs are available at lookup time, price per GPU-hour and billing granularity, CLI/API and agent documentation, signup and minimum-balance friction, built-in time and spend limits, and teardown and data persistence. Source each competitor’s claims from its own documentation.
The Bottom Line
Point the agent at the docs, make it list machines as JSON, launch with both a budget and a TTL, and have it remove the pod when done. Confirm flags with the installed CLI and take prices from the live feed.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
