You kick off a big refactor with your AI coding agent, then you have to leave for lunch. For the next hour you have no idea whether it finished, got stuck, or is sitting there waiting for you to approve a file change. Happy exists for exactly that gap. It lets you watch and steer Claude Code or Codex from your phone while your laptop keeps doing the actual work.
Quick facts
License: MIT. Primary language: TypeScript (about 95 percent of the codebase). Last active: ongoing, with the latest CLI release (1.1.8) shipped in April 2026. It has crossed 21,000 stars on GitHub, which is a lot for a project this young.
What it is
Claude Code and Codex are command-line coding agents. You type instructions, they read your files and write code back. The catch is that they live in a terminal on one machine, so once you step away from that machine, you are cut off.
Happy is a wrapper that fixes this. You install a small command-line tool, then run happy claude instead of claude (or happy codex instead of codex). It starts your agent exactly as before, but now watches the session and mirrors it to an iOS app, an Android app, or a web page. When you pick up your phone, you can read what the agent is doing, reply to it, and approve or deny the permissions it asks for. The moment you touch a key back at your desk, control snaps back to the keyboard.
The whole thing is end-to-end encrypted, which means your code and messages are scrambled on your own devices and only unscrambled on your other devices. The relay server in the middle passes along sealed data it cannot read. The project also ships with no telemetry and no tracking, and since it is open source under the MIT license, you can read the code and confirm that yourself.
The real-world problem it solves
Picture a long-running task: you ask the agent to migrate a database schema and update forty files to match. That runs for a while. Halfway through, it hits a file it is not sure it should touch and pauses for your permission. Without Happy, that pause is invisible until you are back at your desk, and the whole job just sits frozen. With Happy, your phone buzzes, you glance at the request on the train, tap approve, and the migration keeps moving.
Where it's useful
A backend developer starts a slow test-suite fix before a meeting, then clears each permission prompt from their phone under the table so the fix is done by the time they are back.
A solo founder building a side project over the weekend leaves Codex churning on a feature while they walk the dog, checking progress from the sidewalk instead of being chained to the desk.
A freelancer juggling two client codebases spins up separate agent sessions for each and keeps an eye on both from one app, rather than flipping between two terminal windows.
A developer who is privacy-conscious about client code likes that the sync is end-to-end encrypted and that they can self-host the server if they would rather not trust anyone else's at all.
Why it stands out
Two things. First, the handoff is genuinely smooth in the plain sense of the word: you do not "log in" on your phone to take over, you just start typing there, and when you touch your laptop again it takes back over. There is no mode switch to fumble.
Second, the encryption is not a marketing sticker. The design keeps your code readable only on your own devices, and because the whole stack is open, that claim is auditable rather than a promise. For anyone whose day job involves code they are not allowed to leak, that is the difference between "nice idea" and "actually usable."
Worth being honest about the tradeoff: Happy does not replace Claude Code or Codex, it rides on top of them. You still need those agents installed and paid for. Happy adds the mobile eyes and hands, not the brain.
How to get started
You need the agent you already use (Claude Code or Codex) set up on your computer. Then install the Happy CLI:
npm install -g happy
Grab the mobile app from the App Store or Google Play, or just open the web app, and pair it. After that, launch your agent through Happy instead of directly:
# instead of: claude
happy claude
# or
happy codex
If you would rather see it before installing anything, the team has a short demo video.
Takeaway
Happy is a small, focused idea done well: give your terminal coding agent a phone screen and a doorbell. If you have ever walked away from a running agent and spent lunch wondering what it was up to, this closes that loop, and it does it without asking you to hand your code to a server that can read it.