How to Run Claude Code on a Cloud VM

Published July 25, 2026 · by the AQ team

Running Claude Code on a cloud VM takes about fifteen minutes: provision a small Linux server (Ubuntu 22.04 or 24.04 with at least 4 GB of RAM), install the CLI with Anthropic's one-line installer, sign in once over SSH using the paste-code login flow, and run every session inside tmux so it survives disconnects. The payoff is an agent that keeps working after your laptop lid closes, that you can reattach to from any device, and that holds its credentials on a machine you control. This guide walks through the whole setup, hardening included, with every Claude Code detail checked against Anthropic's documentation as of July 2026.

Why put Claude Code on a server at all

On a laptop, Claude Code dies with your terminal: close the lid mid-task and the session is gone. On a VM the process runs on hardware that never sleeps, so a long refactor keeps going through your commute, and you can check on it from a different machine (or your phone with an SSH client) by simply reattaching. Two quieter benefits: the blast radius of an agent running with relaxed permission prompts is confined to a disposable box instead of the machine that holds your email and password manager, and the VM gives agent credentials and repository clones a stable home independent of any laptop.

Step 1: Provision the VM

Anthropic's stated requirements are modest: Ubuntu 20.04+, Debian 10+, or Alpine 3.19+, on x64 or ARM64, with 4 GB or more of RAM. In practice, size for what runs next to the agent. If the repository has a heavy npm or cargo build, or you keep a dev server running while the agent works, 8 GB is the comfortable floor; 2 vCPUs and 4 GB is fine for the agent plus a lightweight toolchain. ARM instances are usually cheaper and work fine, since Claude Code ships ARM64 builds.

Pick Ubuntu 24.04 LTS unless you have a reason not to: every tool in this guide has first-class packages there. On a 4 GB machine, add swap so a hungry build gets slow instead of killed:

sudo fallocate -l 4G /swapfile
sudo chmod 600 /swapfile
sudo mkswap /swapfile
sudo swapon /swapfile
echo '/swapfile none swap sw 0 0' | sudo tee -a /etc/fstab

Any provider works. As of July 2026, a 2 vCPU, 4 GB basic droplet at DigitalOcean runs about $24 per month, Hetzner's shared vCPU plans with the same memory land in the low teens of dollars even after its June 2026 repricing, and AWS Lightsail bundles start at $3.50 per month with flat pricing. Big-cloud on-demand instances (EC2, Google Compute Engine) cost more for the same specs but fold into an existing cloud bill. Prices move; check the provider's pricing page before committing.

Step 2: Ten minutes of hardening

An always-on box with your GitHub credentials deserves basic hygiene before any agent touches it. Create a non-root user, allow SSH through the firewall, and turn off password logins so the internet's constant background brute-forcing hits a wall:

# As root, once
adduser dev
usermod -aG sudo dev

# Firewall: SSH only
ufw allow OpenSSH
ufw enable

# Keys only: in /etc/ssh/sshd_config set
#   PasswordAuthentication no
#   PermitRootLogin no
# then restart sshd

# Security patches apply themselves
apt install -y unattended-upgrades

Two optional upgrades are worth knowing about. A mesh VPN such as Tailscale (which has a free tier for personal use as of July 2026) takes the SSH port off the public internet entirely, so only your enrolled devices can reach the box. And mosh keeps your interactive connection responsive on flaky networks, though with tmux in the loop a dropped SSH connection costs you nothing anyway.

Step 3: Install Claude Code

Install the basics, then Claude Code itself with Anthropic's native installer, which needs no Node.js and updates itself in the background:

sudo apt update && sudo apt install -y git tmux
curl -fsSL https://claude.ai/install.sh | bash

The installer places the binary at ~/.local/bin/claude. Confirm it works, and run the built-in diagnostic if anything looks off:

claude --version
claude doctor

If you prefer everything to arrive through your package manager, Anthropic also publishes signed apt, dnf, and apk repositories with stable and latest channels; those installs update through your normal system upgrades instead of auto-updating.

Step 4: Sign in over SSH

Claude Code requires a Claude Pro, Max, Team, or Enterprise plan, or a Claude Console account with API billing; the free Claude.ai plan does not include it. Authentication is a browser OAuth flow, and it works fine on a headless server: run claude, and when the login screen appears, copy the login URL (press c if it is not shown) and open it in the browser on your laptop. Because the browser cannot reach the callback server on the remote machine, it shows a login code instead; paste that code back into the terminal at the prompt, and you are in. Credentials are stored on the VM at ~/.claude/.credentials.json with owner-only file permissions, so the login survives reboots and you will not repeat this often. Logins do expire eventually; Claude Code warns you a few days ahead, and running /login renews it.

For scripts and CI on machines where nobody can complete a browser hop, Anthropic provides a different tool: claude setup-token mints a one-year OAuth token tied to your subscription, which you export as the CLAUDE_CODE_OAUTH_TOKEN environment variable. For an interactive dev box, the normal login above is simpler and is what you want.

Step 5: Run every session inside tmux

This is the step that makes the whole setup worth it. A Claude Code process launched in a bare SSH session dies when the connection drops. Inside tmux, the session belongs to the server, not to your connection:

# Start a named session per task
tmux new -s auth-refactor
claude

# Detach: press Ctrl-b, then d. Close the laptop. Later, from anywhere:
ssh dev-box
tmux attach -t auth-refactor

# See what is running
tmux ls

Name sessions after tasks, not after the tool. When you run two or three agents at once, tmux ls should read like a to-do list. New windows inside one session (Ctrl-b, then c) work too, but separate named sessions are easier to reason about and to kill cleanly. One warning: tmux state lives in memory, so a VM reboot clears every session. That is rarely a problem for day-long tasks, but it means the box is a workbench, not a system of record; anything worth keeping should be committed and pushed.

If you run several agents in one repository, give each its own git worktree so they do not fight over the checkout; our worktrees guide covers the pattern. For the broader landscape of ways to keep sessions alive, including VS Code tunnels and managed cloud agents, see keeping Claude Code running after closing your laptop.

What this actually costs

The VM is the only new line item. Your Claude usage is billed exactly as before, through the subscription or Console account you logged in with; nothing about running on a server changes model pricing. So the marginal cost of the whole setup is roughly $10 to $25 per month at the providers above, for an agent that works while you sleep.

Where a hand-rolled box runs out of road

For one engineer, the setup above is genuinely good, and you should run it. The friction appears at the edges. Sharing a live session with a teammate means sharing SSH access to the box, which most teams rightly refuse to do. Parallel tasks need per-task worktrees you script yourself. Nobody can see what the agents are doing without logging in and attaching, so a stuck agent stays stuck until you happen to check. And each teammate who wants this needs to repeat the whole ritual on a box of their own.

Where AQ fits

AQ is the multiplayer coding harness where engineering teams run AI coding agents like Claude Code and Codex together: shared live terminals, a code editor, and app previews, in your own cloud. It is this guide's setup, operated as a product. Agents run as real CLIs (Claude Code, Codex, Cursor Agent, Kimi, Grok, or plain shells) in persistent tmux sessions on your team's VM, streamed live to the browser: sessions survive a closed laptop and resume from any device, and a teammate can open the same workspace and see and steer the same live session instead of asking for SSH keys. Each engineer connects their own Claude or OpenAI account, and AQ never marks up model usage.

The glue you would otherwise script is built in: every workspace gets an isolated git worktree on a branch named ai/{id}-{slug} with dependencies installed automatically and a one-click rebase onto main, commits and PRs use per-user GitHub auth, opened PRs are tracked per workspace, and each workspace can run a live dev-server preview with shareable links, where comments pinned on the preview (screenshots included) can be sent to the agent as prompts. Teams that plan in Linear can label an issue ai-task and a workspace appears with status syncing both ways. Runner health is visible in the product, the installer runs preflight checks, and an aq-runner doctor command diagnoses a broken host.

Pricing mirrors the two ways to own the box. The Free plan is the full product on a VM you bring, paying only your cloud provider, exactly like this guide. The Team plan is $100 per user per month (promotional; standard $200, billed monthly) and includes a dedicated always-on VM managed by AQ in its own isolated network, with the rate locked for your first 12 months. Solo and comfortable in tmux? Keep the hand-rolled box. AQ earns its place when the second engineer wants in.