jurniti
Always-on AI agents in their own Firecracker microVMs. BYOK, flat pricing, live in about 3 minutes.
jurniti is a managed platform where each AI agent runs in its own Firecracker microVM with hardware-enforced KVM isolation — the same primitive AWS uses for Lambda — instead of a shared Docker container. Model API keys stay inside your VM (BYOK); jurniti never sees, proxies, or marks up the spend. It supports Claude Code, Codex, Hermes, OpenClaw, Pi and more, with a developer CLI, REST API and MCP server for provisioning VMs, forking behavior templates and running parallel agent swarms. A box is live in about three minutes with an in-browser terminal, persistent disk and your own subdomain.
What is jurniti?
jurniti is a managed agent-hosting platform where each AI agent runs in its own Firecracker microVM — the same isolation primitive AWS uses for Lambda — with hardware-enforced KVM boundaries instead of a shared Docker container. Model API keys are bring-your-own and stay inside the guest VM; jurniti states it never sees, proxies, or marks up model spend. It supports agent runtimes including Claude Code, Codex, Hermes, OpenClaw and Pi, and gives each box an in-browser terminal, persistent disk and TLS subdomain, provisioned in about three minutes. Pricing is flat per VM: Starter $25/mo, Pro $49/mo, Max $99/mo, with no free tier and a 30-day money-back guarantee.
Key features
- One Firecracker microVM per agent with KVM hardware isolation, pinned CPU/RAM cgroup reservations — not a shared-kernel container
- BYOK: model API keys live only inside your VM; jurniti is not a token proxy and adds no markup on model spend
- Managed support for Claude Code, Codex, Hermes, OpenClaw, Pi and many other upstream agent runtimes, installed from upstream installers at first boot
- Developer CLI, REST API and MCP server exposing the same operations (jurniti up / run / cp / fork / templates), mountable into your own agent via 'claude mcp add jurniti'
- Forkable agent templates — a snapshot of a configured VM disk published publicly or shared privately, forked into a new paid VM ('a GitHub for agent behavior')
- Parallel agent swarms: 'jurniti up --tier spot --count N' then drive each box with run/cp, one KVM boundary per agent
- Persistent disk, snapshots/restart, daily backups on Max, public TLS subdomain, in-browser terminal; boots in ~124ms, full provisioning ~3 minutes
- Flat monthly pricing Starter $25 / Pro $49 / Max $99 (annual = 2 months free), 30-day money-back guarantee, no free tier or trial
Who it's for
- Keeping Claude Code, Codex or OpenClaw agents running always-on after the laptop lid closes, with real credentials
- Solo developers and small teams who refuse to put model keys in shared containers and want hardware isolation per agent
- Running parallel agent fleets or swarms where each agent needs its own isolated computer
- Hosting OpenClaw or Hermes with messaging bridges (Discord, WhatsApp, Signal, Telegram) on a persistent box
- Sharing reproducible agent setups by publishing VM-disk templates others can fork
When not to use it
jurniti is not for you if you want a free sandbox or a $4 shared-kernel container — there is no free tier — nor is it a model API, prompt proxy or usage-metered LLM gateway; you bring your own keys. It also does not compile or host your application as a product; the VM is the product and the agent inside it does the work.
FAQ
How much does jurniti cost?
Flat per VM: Starter $25/mo ($250/yr), Pro $49/mo ($490/yr), Max $99/mo ($990/yr); annual billing gives 2 months free. There is no free tier or trial, and every plan carries a 30-day money-back guarantee. Model spend is paid by you directly to the provider.
Is jurniti a real VM or a Docker container?
A real Firecracker microVM per agent with KVM hardware-enforced isolation — the same primitive AWS uses for Lambda — running on AWS EC2 nested-virtualization hosts (us-west-2), with pinned CPU/RAM reservations rather than shared-kernel containers.
Can jurniti see my API keys or prompts?
No. BYOK keys are stored inside your guest VM only; jurniti states it never sees, proxies, or marks up model spend, and it ships upstream agent installers without proprietary patches.
Which agents run on jurniti?
Claude Code, Codex CLI, Hermes (Nous Research), OpenClaw, Pi, OpenCode, Goose, Cline, Kilo Code, Aider and more — each runtime has its own landing page and a canonical list at jurniti.com/llms-full.txt.
Can an agent provision jurniti boxes itself?
Yes. The jurniti binary doubles as an MCP server, so your own agent can call the same operations as the CLI and REST API — mount it with 'claude mcp add jurniti -- jurniti mcp'.
Share this launch
Embed this badge
<a href="https://orangebot.ai/product/jurniti" target="_blank" rel="noopener noreferrer"> <img src="https://orangebot.ai/api/badge/jurniti.svg" alt="Featured on OrangeBot" width="200" height="54" /> </a>