# Claude Code Hosting on a Real microVM > Run Anthropic's Claude Code on its own always-on Firecracker microVM. Log in with your Claude subscription, keep a persistent home, give the agent root without risking your laptop. 124ms boot, BYOK, 30-day money-back. - Published: 2026-06-07 · Updated: 2026-06-07 · jurniti - Canonical: https://www.jurniti.com/blog/claude-code-hosting In October 2025 a developer asked Claude Code to rebuild a project. It ran `rm -rf` starting from `/` and wiped every user-owned file on the machine. That's the trade you make when you give an autonomous agent a shell on the same computer you live on. Here's the fix: run [Claude Code](/claude-code) on its own always-on Firecracker microVM, give it root *there*, and let it work while your laptop stays untouched. The box boots in about 124ms, your Claude login never leaves it, and a bad command stays contained to a machine you can rebuild in seconds. ## What Claude Code is Claude Code is Anthropic's agentic coding tool that lives in your terminal. It reads your codebase, runs commands, edits files, and handles git workflows from natural-language instructions. It's a thin client — the heavy lifting happens on Anthropic's servers — so the thing it actually needs is a place to *run*, with a shell and your files. It's happiest given room to work: hand it a refactor, a failing test suite, or a dependency bump and let it grind. That argues for a box that's always on and holds your project — not a laptop you close at 5pm. ## Why a real microVM for Claude Code A coding agent runs `bash`. It installs packages, edits files, and executes whatever the model decides — code no human reviewed. On your laptop, "whatever the model decides" has the run of your home directory, your SSH keys, and your browser sessions. The October 2025 wipe wasn't a fluke; in January 2026 researchers showed hidden text in a `.docx` could trick Claude Code into exfiltrating files via prompt injection. The safety guidance is blunt, and Anthropic's own docs echo it: before you use bypass mode, isolation is mandatory — a locked-down container or VM. A container shares the host kernel, and there are 200-plus known weaknesses in that stack. **jurniti gives Claude Code its own Firecracker microVM** — its own kernel, a KVM hardware boundary, one tenant per box. The same isolation the serverless platforms use. Now you can hand the agent root and run `--dangerously-skip-permissions` for real autonomy, and the blast radius is one disposable VM. Something goes sideways? Reprovision and you're back in seconds. **You also keep your Claude relationship.** jurniti's [BYOK](/pricing) is architectural, not a setting: run `claude /login` with your Pro or Max plan, or paste an `ANTHROPIC_API_KEY`. The credential lives only in your VM, traffic leaves straight for Anthropic, and jurniti is never in the path. No proxy. No token markup. Because Claude Code is so light, **Starter (1 vCPU / 2 GiB) is plenty** — a VPS plus a Claude plan runs the same money, except here you also get real isolation and zero box ops. And it's reversible: a **30-day money-back guarantee**, workspace kept 7 days after cancellation. ## Run Claude Code in 3 steps By hand this is a VPS, Node, the installer, a non-root user, a systemd unit, tmux so sessions survive a disconnect, and a Docker wrapper if you actually want bypass mode to be safe — then patching it for as long as you run it. jurniti runs the box; you keep the agent. ### 1. Pick a plan and pay [Starter](/pricing) is enough for Claude Code. Check out, and the provision chain runs automatically the moment payment confirms. ### 2. Your microVM boots with Claude Code ready A fresh Ubuntu microVM boots with the Claude Code CLI pre-installed and on your PATH. You get an in-browser terminal — nothing to SSH into. Log in and go: ```bash claude /login # use your Claude Pro/Max subscription # or: export ANTHROPIC_API_KEY=sk-ant-... claude ``` ### 3. Put it to work Drop into a project, hand Claude Code a task, and close your laptop. Your `CLAUDE.md`, slash commands, skills, and subagents live on the persistent home, so they survive every restart. ## What people run Claude Code for - **Overnight work.** Give it a task, close the lid, come back to a finished diff. No dropped session when your WiFi hiccups. - **Full-autonomy runs.** Bypass-permissions mode is genuinely useful when the agent can't escape the box. - **A pinned environment.** Your tools, your `CLAUDE.md`, your MCP servers — set up once, always there. - **Parallel agents.** Run a few microVMs side by side instead of ten terminal tabs you lose track of. Snapshot a dialed-in setup into a [template](/templates) and fork it for the next project — your config, never your credentials. ## jurniti vs a laptop vs a DIY VPS | | jurniti | Your laptop | DIY VPS | |---|---|---|---| | Always on | Yes | No — it sleeps | Yes | | Blast radius of a bad command | One disposable microVM | Your whole machine | Your whole server | | Bypass mode (`--dangerously-skip-permissions`) | Safe — KVM-isolated | Risky | Risky unless you wrap it | | Claude login | Yours, never leaves the VM | Yours | Yours | | Setup | Automatic · 124ms boot | Local install | Manual | | Refund | 30-day money-back | n/a | None | ## Pricing Starter is **$25/mo** ($250/year, two months free) and comfortably runs Claude Code. Pro is **$49/mo** and Max **$99/mo** for heavier workspaces or a custom TLS subdomain. Every plan includes the 30-day money-back guarantee, and your Claude subscription or API spend is separate and never marked up. Running a different agent? See the guides for [OpenAI Codex CLI](/blog/codex-cli-hosting), [OpenCode](/blog/opencode-hosting), and [Hermes Agent](/blog/hermes-agent-hosting), or compare plans on the [pricing page](/pricing). ## Frequently asked questions ### Can I run Claude Code on a server? Yes. Claude Code is a lightweight terminal client — all the model work happens on Anthropic's side — so it runs comfortably on a small always-on box. jurniti gives it a dedicated Firecracker microVM with an in-browser terminal, a persistent home, and isolation from your laptop. You hand it a task, close the lid, and come back to a finished diff. ### Do I need an Anthropic API key, or can I use my Claude subscription? Either. Open the in-browser terminal and run `claude /login` to authenticate with your Claude Pro or Max subscription, or export an ANTHROPIC_API_KEY. jurniti is bring-your-own-key — your credentials live only in your VM, traffic goes straight to Anthropic, and we never see your tokens or mark up your usage. ### Is it safe to run Claude Code with --dangerously-skip-permissions? Far safer on a microVM than on your laptop. The standard guidance — Anthropic's docs included — is that bypass mode belongs in a locked-down container or VM. On jurniti, Claude Code gets root inside its own Firecracker microVM with its own kernel, so a runaway `rm -rf` or a prompt-injection attack is contained to that box — not your machine, and not another tenant's. ### How much does it cost to host Claude Code? Plans start at $25/mo ($250/year — two months free), and every plan has a 30-day money-back guarantee. Claude Code is light on resources, so Starter (1 vCPU / 2 GiB) is plenty. Your Claude subscription or API spend is separate and never marked up, because you bring your own key. ### Will my CLAUDE.md, skills, and slash commands survive a restart? Yes. Your home directory is on a persistent volume, so your CLAUDE.md, custom slash commands, skills, subagents, and MCP config all survive restarts and reprovisions. You can also snapshot the whole setup into a forkable template — credentials are scrubbed on publish. ### Is this an official Anthropic product? No. jurniti is independent managed hosting for Claude Code. We install the upstream CLI at first boot — no fork — so you run the real Claude Code and log in with your own Anthropic account.