Deploy Ubuntu SSH + Claude Code

Deploy and Host an Ubuntu SSH Workstation with Claude Code on Railway

An Ubuntu 24.04 LTS box you reach over SSH or a browser terminal, with Claude Code preinstalled and your home directory on a volume that survives every redeploy. One click, no inputs: the password is generated for you. Node.js 24 LTS, Python 3, git, GitHub CLI, build-essential, tmux and the usual tools are included; dev has passwordless sudo.

Plan requirements. Idle, the box uses about 25 MB of RAM (measured on this image: 22 MB after five idle minutes, 38 MB peak at boot) and is well under Railway's Free and Trial limits. Claude Code sessions add a few hundred MB while they run; builds and language servers add whatever they need. Hobby is the comfortable fit for daily use. The image is pinned by digest, so what you deploy today is what you get on every redeploy.

About Hosting an Ubuntu SSH Workstation

The service runs three small processes: sshd on port 22 behind a Railway TCP proxy, ttyd (a terminal in your browser) behind nginx with basic auth on the public HTTPS domain, and a supervisor that exits if any of them dies so Railway's restart policy takes over. Login is user dev with the generated SSH_PASSWORD; root login is disabled. /home/dev is a Railway volume, so projects, dotfiles, Claude Code's own state (~/.claude), tools you install and even the SSH host keys persist across redeploys. Railway's healthcheck (/healthz) only passes once the terminal actually answers.

Common Use Cases

• Claude Code from anywhere. SSH in from a laptop, or open the browser terminal from a phone or a locked-down machine, and keep an agent session running in tmux while you are away.
• A persistent Linux scratch box. Clone a repo, install what you need, leave it for a week, come back to the same shell.
• A cheap always-on dev environment. Point VS Code Remote-SSH or JetBrains Gateway at it; the box costs cents a day while idle.
• Long-running jobs without a laptop. Data pulls, builds and scripts that outlive a local terminal session.

Dependencies for Ubuntu SSH Workstation Hosting

None. A single service with one volume. Claude Code needs an Anthropic account: an API key or a Claude subscription login, both optional at deploy time.

Deployment Dependencies

• Image source and runtime contract
• Claude Code documentation
• ttyd (browser terminal)
• Railway TCP proxying
• Railway volumes

Implementation Details

Connecting over SSH. Open the service, then Settings → Networking → TCP Proxy: it shows a host and port. Connect with

ssh dev@your-service.proxy.rlwy.net -p 12345

The password is SSH_PASSWORD in the service's Variables tab (Railway generated it at deploy). The deploy log also prints the exact ssh line once at boot. Host keys are stored on the volume, so a redeploy will not trigger a host-key-changed warning.

Connecting from the browser. Open the service's public URL. Sign in with user dev and the same SSH_PASSWORD. You land in a tmux session named main: close the tab and reopen it later and the session, and anything running in it, is still there. Works from a phone.

Adding SSH keys. Either paste public keys into the AUTHORIZED_KEYS variable (one per line) and redeploy, or, once you are in, append them to ~/.ssh/authorized_keys on the volume. Password login stays enabled; to turn it off, add PasswordAuthentication no to /etc/ssh/sshd_config.d/20-local.conf — note that this file lives in the image layer and is reset on redeploy, so keep it in the volume and symlink it from ~/.bashrc if you need it permanent, or just use a long generated password.

Claude Code. Run claude. With no key set it offers the Claude subscription login: open the link it prints on any device, paste the code back. The login is stored in ~/.claude on the volume, so you do it once. To use an API key instead, set ANTHROPIC_API_KEY in the Variables tab and redeploy. The pinned copy in the image is Claude Code 2.1.278; to move to the latest, run curl -fsSL https://claude.ai/install.sh | bash — it installs into ~/.local/bin on your volume, which is first on PATH.

Codex, opencode and other tools. npm install -g @openai/codex or npm install -g opencode-ai install into ~/.npm-global on the volume without sudo. pip, pipx, uv, cargo and anything under ~/.local persist the same way. For apt packages use sudo apt-get install; those live in the container layer and are gone after a redeploy, so put the command in a script on the volume if you rely on them.

GitHub. Set GITHUB_TOKEN and both git (HTTPS clone/push to github.com) and gh authenticate with it. Or run gh auth login interactively; its state persists on the volume too.

Environment variables. Every variable you add in the Variables tab is exported into your SSH and browser sessions after the next deploy, except SSH_PASSWORD. That is how ANTHROPIC_API_KEY, GITHUB_TOKEN, TZ and anything else of your own reach the shell.

What persists and what does not. Persists: everything under /home/dev (projects, dotfiles, ~/.claude, ~/.local, ~/.npm-global, ~/.ssh), and the SSH host keys. Does not persist: apt packages, files outside /home/dev, and running processes (a redeploy restarts the container, so tmux sessions end). Redeploying a service with a volume involves a few seconds of downtime; that is a Railway volume property, not a template setting.

Rotating the password. Change SSH_PASSWORD in the Variables tab; the next deploy applies it to SSH and the browser terminal together.

Cost. Railway bills RAM and CPU by the minute. Measured on this image: 22 MB idle, which is about $0.25/month at Railway's published rates before any work you run. Claude Code and builds add usage while active. There is nothing here that keeps the box busy on its own.

Security notes. The box is a real Linux machine on the public internet: SSH is exposed through Railway's TCP proxy (with a 24-character generated password, root login disabled, four tries per connection) and the browser terminal through HTTPS basic auth. Keys are better than passwords; add yours. dev has passwordless sudo inside the container, which is the point of a dev box, not a hole in one. This template is for personal development use; running it as a shared shell service or reselling access is outside Railway's Fair Use policy.

Why Deploy an Ubuntu SSH Workstation on Railway?

Railway is a singular platform to deploy your infrastructure stack. Railway will host your infrastructure so you don't have to deal with configuration, while allowing you to vertically and horizontally scale it.

By deploying an Ubuntu SSH Workstation on Railway, you are one step closer to supporting a complete full-stack application with minimal burden. Host your servers, databases, AI agents, and more on Railway.