Happy integrates models, teams, and compute into one secure, open-source harness — accessible from terminal, desktop, and mobile, deployable anywhere, and adaptable to your team.
🖥️ Download for macOS • 📱 iOS App • 🤖 Android App • 📚 Documentation • 💬 Discord
export-1787511700238.mp4
Open it — setup runs itself.
Scan the QR code in the desktop app — your sessions follow you everywhere.
Open a project and ask for what you want. Invite your team when you want company.
Serious work with coding agents outgrows a terminal tab fast. Sessions die with the window that held them. Every vendor's agent lives in its own app with its own tools, its own permissions, and its own idea of a session, so switching models means abandoning context. Your teammates can't see what your agent is doing, let alone step in. And the moment the work matters, someone asks where the code is going and who can read it.
Happy exists to remove those walls. One harness runs every agent you already pay for, keeps every session alive and durable, lets your whole team work inside the same conversation, and never ships your work anywhere you didn't point it. This repository is the macOS app — the place where that harness becomes a full working environment: conversations beside the files, diffs, terminals, and previews the work actually touches.
An agent session shouldn't be a private transcript. Bring your team into one session with every agent: anyone can share context, steer the conversation, approve decisions, and take over in real time. Start a task at your desk, check it from your phone, hand it to a teammate — it's the same session, the same permission boundary, the same history, on every device.
The best model for planning isn't the best model for building or reviewing, and you shouldn't have to abandon your context to switch. Let Claude plan, Codex build, and Grok review — or run them side by side and compare. Each model still gets its native prompts and tools, so nothing is dumbed down to a lowest common denominator, but the session, the permissions, and the context stay together across every handoff. Happy adds no account of its own; it uses the sign-ins already on your machine.
A tool this close to your code has to answer to you, not to a vendor's roadmap. Happy is open source end to end: run it on your hardware, in your cloud, or in ours, and change it to fit how your team actually works. Happy adopts itself to build itself — the pressure of daily use is the product roadmap.
You can't bring an agent into real work if you can't say where the data goes. With Happy the answer is short: nowhere you didn't send it. No telemetry, no hosted account, no third-party servers by default. Your projects stay ordinary directories on your machine, and everything that travels between agents, teammates, and devices is end-to-end encrypted — relays carry ciphertext they cannot read. That is what makes Happy safe to run inside a corporate network: the only party that ever sees your prompts is the model provider you chose, under its own terms.
Download the desktop app and open it — that is the whole setup. Happy downloads and starts its agent runtime on its own, finds the Claude, Codex, and Grok sign-ins already on your machine, and asks you to point it at a folder you work in. From there, just work: open a project, start a session, and ask for what you want.
When you want company, invite people. Share a session with a teammate and you are both inside the same conversation — steering, approving, and taking over in real time — or grab the mobile app and take your own sessions with you.
- Projects, workspaces, and sessions. Navigate everything the daemon knows about, start new sessions where the work belongs, and keep parallel efforts organized instead of scattered across terminal tabs.
- Durable, streaming conversations. Watch an agent think and act in real time; the transcript is a permanent record, not a scrollback buffer.
- Files, diffs, and editing. See what changed, review diffs, open and edit files, and preview results without leaving the app.
- Terminals. Real terminals attached to the machine doing the work.
- Browser and HTML preview. Open web content and rendered HTML inside Happy instead of bouncing to an external browser.
- Models, providers, and usage. Choose models and effort per session, and see what your work is costing as it happens.
- Notes and visibility. Keep notes alongside the work, and see what agents and their processes are doing at any moment.
We're engineers scattered across Bay Area coffee shops and hacker houses, constantly checking how our AI coding agents are progressing on our pet projects during lunch breaks. Happy was born from the frustration of not being able to peek at our AI coding tools building our side hustles while we're away from our keyboards. We believe the best tools come from scratching your own itch and sharing with the community.
- Documentation Website - Learn how to use Happy effectively
- Development Guide - Repository setup, browser mode, profiling, and validation
- Discord - Ask questions and share what you build
MIT License - see LICENSE for details.