Docs / Guides
Compute for AI agents: servers that LLMs can rent themselves
Give your AI agent its own infrastructure: an API-first VPS platform with llms.txt, crypto payments and hourly billing.
Agents need disposable compute
An autonomous agent that can write code eventually needs somewhere to run it: a sandbox for risky scripts, a browser farm for research, a runtime for tools it builds for itself. Giving it your laptop is a bad idea; giving it a disposable VPS with a hard budget is a good one.
CryptoVPS is API-first with no human-only steps: registration is an email, payment is a balance, deploys are a single POST. Our llms.txt gives any agent — Gemini, GPT, Claude or your own — the full instructions in one fetch.
The budget is the leash
Fund the balance with exactly what you are willing to spend. The agent can deploy at most 10 servers, each costs cents per hour, and when the balance ends everything suspends — a natural spending cap with no surprise invoice at the end of the month.
Patterns that work
Ephemeral sandboxes (deploy, run untrusted code, destroy), scheduled workers the agent re-creates on demand, and self-hosted tools the agent installs for itself. The API reference at /docs/api covers polling, error handling and the rules agents should follow.