π GitHub β’ π₯οΈ Web App β’ π Documentation β’ π¨π³ δΈζ
npm install -g happy-next-cli# Instead of: claude
# Use: happy
happy
# Instead of: codex
# Use: happy codex
happy codex
# Instead of: gemini
# Use: happy gemini
happy gemini
Running happy prints a QR code for device pairing.
- Open
https://happy.hitosea.com/and scan the QR code (or follow the link shown in your terminal). - Prerequisite: install the vendor CLI(s) you want to control (
claude,codex, and/orgemini).
Happy Next works with the hosted server by default (https://api.happy.hitosea.com) and the hosted web app at https://happy.hitosea.com/.
If you want to self-host:
cp .env.example .env
# edit .env
docker-compose up -dNote: this default stack also starts happy-voice. You must configure LiveKit + provider keys (OpenAI/Cartesia/etc.) in .env. See docs/self-host.md.
First run only (apply DB migrations):
docker-compose exec happy-server yarn --cwd packages/happy-server prisma migrate deployOpen the web app at http://localhost:3030.
Full guide: docs/self-host.md
To point the CLI at your self-hosted API:
HAPPY_SERVER_URL=http://localhost:3031 HAPPY_WEBAPP_URL=http://localhost:3030 happyHappy Next intentionally changed client KDF labels as part of the rebrand. Treat this as a new generation: do not expect encrypted data created by older clients to be readable by Happy Next (and vice versa).
Happy Next is a major evolution of the original Happy. Here are the highlights:
- All three agents are first-class citizens with session resume, duplicate/fork, and history
- Multi-agent history page with per-provider tabs
- Per-agent model selection, cost tracking, and context window display
- ACP and App-Server (JSON-RPC) backends for Codex
- AI backend profiles with presets for DeepSeek, Z.AI, OpenAI, Azure, and Google AI
- LiveKit-based voice gateway with pluggable STT/LLM/TTS providers
- Microphone mute, voice message send confirmation, "thinking" indicator
- Context-aware voice: app state is injected into the voice LLM automatically
- Auto-switch providers by prefix (e.g.
openai/gpt-4.1-mini,cartesia/sonic-3)
- Create, switch, and archive multi-repo workspaces from the app
- Per-repo branch selection, settings, and scripts
- Aggregated git status across repos
- Auto-generate workspace
CLAUDE.md/AGENTS.mdwith@importrefs - Worktree merge and PR creation with target branch selection
- AI-powered PR code review with results posted as GitHub comments
- Full file browser with search, Monaco editor viewing/editing
- Commit history with branch selector (local + remote)
- Git changes page: stage, unstage, commit, discard
- Per-file diff stats (+N/-N) for Claude, Codex, and Gemini
- Share sessions with friends via direct invite or public link
- End-to-end encrypted: NaCl Box (direct) and token-derived keys (public links)
- Real-time sync of messages, git status, and voice chat across shared users
- Access control with view, edit, and admin permission levels
- "All / Shared with me" filter tabs and share indicator in session list
- Public share web viewer for link-based access
- Connect to external AI machines via relay tunnel or direct WebSocket
- Machine pairing with Ed25519 key exchange
- Chat interface with real-time streaming and session management
- Task list with filters, search, pagination, and status workflows
- Task detail with HTML rendering, assignees, files, sub-tasks
- Real-time WebSocket chat (Slack-style layout, emoji reactions, voice playback, images/video)
- One-click AI session launch from any task (MCP server passthrough)
- Create tasks and projects directly from the app with cross-platform date picker
- One-command
docker-compose up(Web + API + Voice + Postgres + Redis + MinIO) - Separate origins architecture (no path reverse proxy)
.env.examplewith full configuration reference- Runtime env var injection for Docker builds
- v3 messages API with seq-based sync, batch writes, and cursor pagination
- HTTP outbox for reliable delivery when WebSocket is unavailable
- Server-confirmed message sending with retry
- Fixes for cursor skip, outbox race, message duplication/loss
- Image attachment and clipboard paste (web)
/duplicatecommand to fork a session from any message- Message pagination, unread blue dot indicator, compact list view
- Session rename with lock (prevent AI auto-update), search in history
- Options click-to-send / long-press-to-fill, scroll-to-bottom button
- Pull-to-refresh, inset dividers, Agent tool display with robot icon
- Tool input/output formatted as key-value pairs instead of raw JSON
- In-memory SWR cache and search for agent session history
- CLI:
happy updateself-upgrade,happy --versionwith all agent versions
- 200+ bug fixes: message sending reliability, session lifecycle, Markdown rendering, navigation, voice, DooTask, sharing
- Security: shell command injection fix, plan mode permission handling
- Performance: payload trimming for mobile, lazy-load diffs, rendering optimization
- Dark mode fixes throughout the app
- i18n improvements (Chinese Simplified/Traditional, CJK input handling)
- Markdown rendering: tables, inline code, nested fences, clickable file paths
- Keyboard handling, loading states, navigation stability
Full changelog: docs/changes-from-happy.md
On your computer, run happy instead of claude, happy codex instead of codex, or happy gemini instead of gemini to start your AI through our wrapper. When you want to control your coding agent from your phone, it restarts the session in remote mode. To switch back to your computer, just press any key on your keyboard.
- ποΈ Remote control for Claude, Codex & Gemini - All three agents as first-class citizens
- β‘ Instant device handoff - Take back control with a single keypress
- π Push notifications - Know when your agent needs attention
- π E2EE + self-host option - Encrypted by default, one-command Docker deployment
- ποΈ Voice assistant - LiveKit-based voice gateway with pluggable STT/LLM/TTS providers
- π§° Multi-repo workspaces - Worktree-based multi-repo flows with branch selection and PR creation
- π Code browser & git management - Browse files, view diffs, stage/commit/discard from your phone
- π DooTask integration - Task management with real-time chat and one-click AI sessions
- Happy App - Web UI + mobile client (Expo)
- Happy CLI - Command-line interface for Claude Code, Codex, and Gemini
- Happy Server - Backend server for encrypted sync
- Happy Voice - Voice gateway (LiveKit-based)
- Happy Wire - Shared wire types and schemas
We build Happy Next because we want to supervise coding agents from anywhere (web/mobile) without giving up control, privacy, or the option to self-host.
- Documentation - Learn how Happy Next works (protocol, deployment, self-host, architecture)
- CONTRIBUTING.md - Development setup and contributing guidelines
- SECURITY.md - Security vulnerability reporting policy
- SUPPORT.md - Support and troubleshooting
MIT License - see LICENSE for details.


