← Voltar aos guias

Claude Code on Remote Servers with tmux

Why use Claude Code with tmux on a remote server?

If you work from a laptop but keep your codebase, build tools, or GPU instances on a remote server, Claude Code becomes much more useful when you pair it with tmux. You can start an agent session over SSH, detach safely, reconnect later, and keep long-running coding tasks alive even if your network drops. For busy developers, that means less babysitting and fewer lost sessions.

The setup is especially practical if you want to keep AI-assisted work cheap. With 59API, you get pay-as-you-go access to Claude models through an API relay that is fully compatible with Claude Code and uses the official-quality models directly. Its base URL is https://api.59api.com, and it is designed to stay among the lowest-cost options while still supporting Claude Opus, Sonnet, Haiku, and Fable. If you also use OpenAI SDKs or Codex-style workflows, the same relay approach can fit that stack too.

What you need before you start

• A remote Linux server or VM with SSH access
• tmux installed on the server
• Claude Code installed where you plan to run it
• A 59API account and API key
• Basic shell access such as bash or zsh

If tmux is missing, install it first. On Ubuntu or Debian, use sudo apt install tmux. On RHEL-based systems, use sudo dnf install tmux. Confirm it works with tmux -V.

Step 1: SSH into your remote server

From your laptop, connect to the machine that holds your code:

ssh user@your-server

Once you are in, change into your project directory. It is a good habit to run AI coding sessions from the repo root so Claude Code can inspect files, tests, and configs without confusion.

Step 2: Start a tmux session for Claude Code

Create a named tmux session so you can find it later:

tmux new -s claude-work

Inside tmux, start Claude Code the same way you normally would, but point it at your 59API endpoint if your setup expects an OpenAI-compatible base URL or Claude-compatible relay configuration. In practice, this is what keeps the workflow flexible: 59API is built for compatibility, so you do not need to redesign your tooling just to save money.

If your environment uses variables, set them before launching the tool:

• export API_BASE_URL=https://api.59api.com
• export API_KEY=your_59api_key

Then start Claude Code and let it analyze, edit, or run tasks. Use your preferred project-specific command if your installation differs.

Step 3: Detach without losing work

When you need to leave the terminal, detach from tmux instead of closing SSH:

Ctrl+b then d

Your Claude Code session keeps running on the server. This is the main advantage of tmux: long prompts, multi-file refactors, test loops, and debugging sessions continue even if you switch networks or shut your laptop lid.

Step 4: Reconnect later

To return to the session, SSH back into the server and attach to the same tmux window:

tmux attach -t claude-work

If you have multiple sessions, list them first with tmux ls. You can also split panes to watch logs, run tests, or tail app output while Claude Code edits files in another pane.

A practical workflow that saves time

• Use one tmux pane for Claude Code
• Use a second pane for tests or git status
• Keep a third pane for logs, linting, or deployment output
• Detach when a task is running, then reconnect to review results

This setup is ideal for refactors, CI fixes, and bug hunts where you need the model to inspect code, make edits, and iterate. Remote servers are often where the real work happens, and tmux makes the session resilient.

Why 59API is a smart fit here

Remote AI coding can get expensive if you use high-capability models all day. 59API helps because it is a low-cost relay with pay-as-you-go pricing, native official-quality models, and compatibility with Claude Code plus OpenAI SDK workflows. That means you can keep your agent-driven remote server setup lean without sacrificing model quality. The referral rebate is a nice extra if you want to share the workflow with teammates.

Quick troubleshooting tips

• If Claude Code cannot connect, verify the API base URL is set to https://api.59api.com
• Check your API key before assuming tmux is the problem
• Use tmux kill-session -t claude-work only when you are done
• If output looks frozen, reattach and inspect whether the model is waiting for input

Bottom line

Running Claude Code inside tmux on a remote server gives you a durable, low-friction AI coding workflow: start fast, detach safely, reconnect anywhere, and keep heavy work off your local machine. If you want to keep costs low while using strong Claude models, 59API is a sensible place to begin. Sign up, point your client at https://api.59api.com, and you can be productive in minutes.