Skip to content
Get packs

Guide · Open source

How to use Happy: install the CLI and run Claude or Codex

Updated · written with AI from the project's own documentation and checked against our catalogue data

Happy is an open-source (MIT) project from slopus that lets you run AI coding agents such as Claude Code and Codex, and watch or steer them from a phone. This guide is for developers who already pay for one of these subscriptions and want a simple way to use them. By the end you will know how to install the original Happy CLI, start a Claude Code or Codex session, and find the other parts of the project.

Happyslopus/happy · ★ 24.4k · MIT

Quick start

  1. 1Check prerequisites. Confirm you have Node.js installed and an existing Claude Code or Codex installation or subscription.
  2. 2Install the CLI. Run npm install -g happy in your terminal to install the original Happy CLI globally.
  3. 3Start Claude Code. Run happy claude from your project folder to start a Claude Code session through Happy.
  4. 4Or start Codex. Run happy codex instead if you want a Codex session.
  5. 5Continue on your phone. Open the Happy mobile app to pick up the session from your phone, using the App Store or Google Play listing.
  6. 6Move to the desktop app for more. Use the desktop app and Happy Harness for multi-model sessions and shared sessions with others.

What Happy is and what it reuses

Happy is a set of apps and tools around AI coding agents. According to its README, you sign in with the subscriptions you already pay for rather than buying new access. Happy is not a separate model.

The project is split across several repositories: a mobile app (iOS, Android and web), the original CLI, and a relay server for encrypted sync. A desktop app for macOS, Windows and Linux lives in its own repository.

  • Original CLI: wraps the Claude Code and Codex harnesses in your terminal.
  • Mobile app: lets you continue a session from your phone.
  • Desktop app: the README points to it for multi-model sessions and teams.
  • Relay server: handles end-to-end encrypted sync between your devices.

What you need before you install

The README lists two prerequisites for the original CLI: Node.js and an existing Claude Code or Codex installation or subscription. It does not state which Node.js version is required, so check the project documentation before you start.

  • Node.js (the README does not specify a version)
  • An existing Claude Code or Codex installation or subscription

Install the original Happy CLI

The README gives one install command for the original CLI. It installs the happy command globally through npm, so you can run it from any folder.

npm install -g happy

Start a Claude Code or Codex session

Once installed, start a session with the command for the agent you want. The CLI wraps that agent directly in your terminal, so your work stays in an ordinary project folder.

  • Use happy claude to start Claude Code.
  • Use happy codex to start Codex.
# Start Claude Code happy claude # Or start Codex happy codex

Continue a session on your phone

The README says the original CLI can continue sessions on your phone. Happy's mobile app is listed on the App Store and Google Play, and a web client is also available. Sessions sync between devices through the relay server, and the README says what moves between your devices is encrypted.

For multi-model sessions, inviting others into a session, or running Gemini and other agents, the README directs you to the desktop app and Happy Harness rather than the original CLI.

Know when to use the original CLI and when to use the desktop app

The original CLI is the most direct option if you want one model and a simple terminal command. The README says the multi-model features point to the desktop app and Happy Harness.

Happy's other stated features include subagents on any supported model, and the ability to invite a colleague into a session so you both watch and steer the same agent. Check the documentation at happy.engineering for current setup details.

  • Single model in a terminal: original CLI.
  • Several models, subagents or shared sessions: desktop app and Happy Harness.
  • Reading or forking the code: the repository is MIT licensed.

What people use it for

  • Start a Claude Code session in your terminal and check its progress from your phone while you are away from your desk.
  • Run Codex through the same CLI when you want a second agent for a task.
  • Invite a colleague or friend into a session so you both can watch the agent work and steer it.
  • Read the MIT-licensed code to understand how the wrapper works, or fork it to build your own version.

Frequently asked questions

Is Happy free to use?

Happy's code is released under the MIT licence, so you can read, fork and ship your own build. The README says you sign in with the Claude, Codex or Grok subscriptions you already pay for, so you still need those subscriptions to use the agents.

What do I need to install Happy?

The README lists Node.js and an existing Claude Code or Codex installation or subscription as prerequisites. It does not specify a Node.js version, so check the project documentation for that detail.

How do I install Happy?

The README gives the command npm install -g happy to install the original CLI globally. After that you can start a session with happy claude or happy codex.

Is Happy secure enough for production work?

The README describes end-to-end encryption of what moves between your devices, but it does not describe a security review or a threat model. If you need an audited, production-grade security tool, check for that evidence before relying on it.

How does Happy compare to Claude Code, Cline, Aider or OpenHands?

Claude Code is one of the agents Happy wraps, so Happy adds a phone and multi-agent layer on top of it. Cline, Aider and OpenHands are separate coding agents with their own workflows; the README does not compare them directly with Happy.

Can I use Happy without a Claude or Codex subscription?

The README centres on reusing existing logins, and it lists an existing Claude Code or Codex installation or subscription as a prerequisite. Without one of these, the README does not describe another way in.

New agent skills, repos and AI tools — every week

The best new skills and open-source projects, the fastest-growing repos, and one free prompt. Five minutes to read. Unsubscribe anytime.

Our verdict

Happy is a good fit if you already pay for Claude, Codex or Grok and want one interface, phone access and an MIT-licensed codebase you can read or fork. Pick something else if you need a security-audited tool, or if you only want a single model in a plain terminal, where Claude Code or Aider may be more direct.

Install options & downloads