Free tool
Don’t self-host an AI agent cold. Get a deploy recipe.
jurniti runs Claude Code, Codex, Hermes, OpenClaw, and more — each in its own Firecracker microVM. This tool does the selection work you’d otherwise do before renting a VPS: which agent first, what to add next, plan floors, and what keys you paste at boot. Then start checkout for the first box. No free tier — 30-day money-back instead.
Your deploy recipe
Answer a few agent-job questions. We pick the harness stack, plan floors, and what to paste at first boot — then you start checkout for the first microVM. No free trial; 30-day money-back on every plan.
Start with Claude Code on a Firecracker microVM.
Recommended stack
- AgentClaude Code · from $25/moMatches your model home for terminal-first coding work.
- Start with
Claude Code
from $25/mo
Matches your model home for terminal-first coding work.
First boot: Anthropic API key or Claude subscription token for the CLI
Claude Code product page
Stack for start + next: about $25/mo across 1 microVM (each is its own subscription). Model tokens are BYOK — never marked up.
- Each harness is a separate paid microVM — start with the first, expand after it is live.
- Paste keys only inside your VM (BYOK). jurniti does not proxy or mark up model tokens.
- Every plan includes a 30-day money-back guarantee on first purchase.
First ship: List the tools you have in this environment, then propose a 3-step plan to ship a tiny hello-world change and open a PR.
How this works
- Answer job, model home, always-on, multi-agent (under a minute).
- Copy your recipe — harness order, plan floors, first-boot keys, first task after deploy.
- Deploy the start-with microVM via checkout. Add the next agents after it is live.
Pair with the DIY vs managed cost calculator or the multi-agent org chart.
FAQ
Is this free AI agent hosting?
No. The tool is free. Runtime is paid Firecracker microVMs with a 30-day money-back guarantee — no free trial tier.
Which harnesses can the recipe recommend?
Coding CLIs (Claude Code, Codex, Grok Build, …), always-on agents (Hermes), messaging bridges (OpenClaw), multi-agent control planes (Paperclip), and more — each as its own microVM.
Should I self-host an AI agent on a VPS or go managed?
Self-hosting (an OpenClaw VPS, a Claude Code server you patch yourself) is real ongoing work: updates, key hygiene, uptime. The recipe gives you the same stack either way; the cost calculator compares the two honestly — including when DIY is cheaper.
Do I have to give my email to get my recipe?
No. The tool and its full result are free with no email. Leaving an email is optional — it sends you your deploy plan and saves it for checkout.
Do you mark up model tokens?
No. BYOK inside the guest. Hosting is the flat plan sticker.