Introduction

Remotehost gives every coding agent its own isolated development environment. Run Claude Code, Codex, or both, each in a fresh sandbox with its own processes, ports, package cache, environment variables, and filesystem — so parallel agents never step on each other or on your machine.

The core idea

Coding agents run fastest when they can take real actions — install packages, run tests, start servers — without asking permission for every step. Doing that safely on your laptop means one agent at a time, and a workspace that's slowly accumulating agent-generated mess.

Remotehost solves this by giving each agent a disposable, full-fidelity Linux host instead:

rh claude
rh codex
rh

rh claude and rh codex create an isolated sandbox for the current repository and launch the agent inside it over SSH. Running rh with no arguments opens an interactive terminal UI for observing and managing the sandboxes, orgs, and projects you already have.

Remotehost stays narrow on purpose — it is not a workflow builder, CI/CD replacement, or code review tool. Claude Code and Codex do the coding; Remotehost provides the machine they run on.

What's in these docs

• Getting Started — install the CLI, log in, and launch your first sandbox.
• Concepts — how orgs, projects, sandboxes, and machine sizes fit together.
• CLI Reference — every rh command, its flags, and what it does.
• Console — managing hosts, team members, and billing from the web console.
• Pricing & Billing — how usage is metered and what each plan includes.
• API Reference — the REST API behind the CLI and console.