Happy

Happy is a harness for AI coding agents. Instead of juggling separate interfaces for Claude, Codex, and Grok, you get one app that connects to all three using the subscriptions you already have. No new API costs, no separate billing layer.

The core idea is straightforward: pick the model that fits the task, run it, and let Happy manage the session. You can delegate work to subagents across different providers from the same interface, which matters when one model handles a task better than another.

Key capabilities:

• Multi-provider orchestration connects Claude, Codex, and Grok in one place, each using its vendor's native prompts and tooling
• Subscription reuse means you sign in with your existing Claude, Codex, and Grok accounts rather than paying for another service on top
• End-to-end encrypted mobile app lets you resume or start sessions from your phone when you're away from your desk, with encrypted sync between devices
• Team server support makes it natively multiplayer: deploy a shared server, and everyone authenticates as themselves while sharing projects and sessions
• Terminal compatibility keeps the CLI-first workflow intact; start Claude Code or Codex in your terminal and pick up that session anywhere

Projects stay as ordinary folders on your own hardware. The code is MIT-licensed, so you can read it, fork it, or ship a custom build. For teams that want a conductor-style layer over multiple AI agents without locking into a managed platform, Happy handles that without adding another subscription to the pile.