happier-dev/happier: Cross-Device Control for AI Coding Agents

happier-dev/happier: Cross-Device Control for AI Coding Agents

AI coding agents like Claude Code, Codex, and OpenCode have transformed how developers write software—but they come with a frustrating limitation: you're tethered to your terminal. Start a complex refactoring session on your laptop, step away for coffee, and you're blind to progress. Need to approve a permission request while away from your desk? You're stuck. happier-dev/happier solves this by acting as a secure bridge between your local development environment and every device you own.

This open-source tool lets you run AI coding sessions locally, then monitor, control, and continue them from your phone, browser, or desktop app—without losing context. With 1,349 GitHub stars, 109 forks, and an MIT License, it's gaining traction among developers who need genuine cross-device continuity for agentic coding workflows. Built in TypeScript with end-to-end encryption via TweetNaCl, it prioritizes privacy while delivering the multi-device experience that native CLI tools simply don't provide.

What is happier-dev/happier?

happier-dev/happier is an open-source, end-to-end encrypted, cross-device companion application and client for AI coding agents. It sits alongside tools like Claude Code, OpenAI Codex, OpenCode, Gemini, Kimi, Qwen, and Augment Code—providing a unified interface to start, monitor, and control sessions across multiple devices.

The project emerged from real developer necessity. The maintainers were active contributors to Happy, submitting fixes and improvements upstream. When their own workflow demands outpaced what they could explore comfortably in the main project, they built Happier for internal use first—then open-sourced it after weeks of refinement. The codebase remains deeply grateful to its predecessor: "Happier would not exist without it."

Technically, Happier operates through three core components: a relay server (self-hostable or cloud-hosted at api.happier.dev), a machine daemon that manages local LLM processes, and UI clients (native mobile, web, desktop). All communication flows through the relay, with sessions encrypted on-device before transmission. The architecture supports multiple servers simultaneously—Happier Cloud, personal self-host, and company self-host—with isolated daemon state per server profile.

The project is in alpha preview as of its last commit (2026-07-16), with rapid iteration ongoing. It's developed in Switzerland with explicit focus on data protection and transparency—not the typical "move fast, encrypt later" approach common in the AI tooling space.

Key Features

Broad provider support distinguishes Happier from single-vendor solutions. It unifies Claude Code, Codex, OpenCode, Gemini, GitHub Copilot, Kiro, Pi, Kilo, Kimi, Qwen, Augment, and any custom ACP-compatible CLI behind one interface. You don't need separate workflows for each agent.

Session continuity is the core value proposition. Start locally, continue remotely. The system supports browsing existing sessions, real-time following of live sessions started outside the app, and full takeover with context preservation. Fork sessions at any message; handoff live sessions between machines while maintaining the same session ID.

Collaborative infrastructure enables sharing live sessions via view-only public links or direct username invites. Teammates can join sessions, with pending queues for messages while agents are busy—editable, reorderable, and synchronized across collaborators.

Voice integration goes beyond basic speech-to-text. The voice agent monitors all running sessions, switches focus between them, reads and answers permission requests, sends dictated messages, and discusses agent activity with full context access. It runs on daemon-backed AI (Claude or configurable backends) with ElevenLabs realtime, Kokoro neural TTS, or device/Google STT options.

Enterprise hardening includes GitHub OAuth with org/team gating, OIDC (Okta, etc.), mTLS with certificate identity mapping, and keyless external auth. Storage policies range from end-to-end encrypted (default) to plaintext-only for organizations managing encryption at infrastructure level. Server owners can disable capabilities via environment variables, with UI adapting automatically.

Additional capabilities include: MCP server definitions reusable across all providers; persistent projects and worktrees outside sessions; embedded terminal dockable in multiple layouts; local memory search across decrypted transcripts; and smart notification routing that opens exact sessions needing attention.

Use Cases

Remote session monitoring solves the "coffee break problem." Start a long-running Codex refactoring session, leave your desk, and approve permission requests or inject steering messages from your phone. The inbox aggregates all pending actions across sessions and machines—no context switching between apps.

Multi-machine development suits developers with separate work and personal machines, or those using cloud instances. Start a session on your desktop, handoff to your laptop for a meeting, attach via happier attach from any terminal. The session ID persists; only the active machine changes.

Team pair programming leverages collaborative sessions with fork-and-replay. One developer starts a Claude Code session exploring an architecture problem, shares the link, and teammates can follow live or fork at any point to explore alternatives. The pending queue lets multiple people queue suggestions without talking over each other.

Voice-driven oversight benefits developers who need eyes-free interaction. The voice agent can monitor multiple sessions simultaneously, alert you to permission requests, and execute approved actions—all while you're reviewing code on another screen or away from your keyboard. Every action maps to an individually approval-gatable Happier action.

Enterprise self-hosting addresses organizations blocked by SaaS-only AI tools. Run the relay server internally, enforce GitHub org membership, configure mTLS, set session retention policies, and disable features like social sharing or bug report uploads per compliance requirements. The Docker^{↗ Bright Coding Blog} deployment supports PostgreSQL^{↗ Bright Coding Blog}, SQLite, or MySQL^{↗ Bright Coding Blog} backends.

Installation & Setup

Happier requires installing both the CLI/daemon on your development machine and at least one client app.

Step 1: Install the CLI

macOS/Linux:

curl -fsSL https://happier.dev/install | bash

Windows:

iwr https://happier.dev/install.ps1 -useb | iex

For npm specifically:

npm install -g @happier-dev/cli

Step 2: Authenticate

happier auth login

The maintainers recommend mobile-first authentication—your account and secret key are then securely stored on your mobile device. If you open a terminal-connect link while logged out, Happier routes you through sign-in before returning to approval.

Step 3: Start using Happier as your agent wrapper

# Replaces: claude
happier

# Replaces: codex
happier codex

# Replaces: opencode
happier opencode

# Additional providers
happier gemini
happier kilo
happier kimi
happier qwen

Client apps

• iOS: Available on the App Store
• Android: Private beta via Google Group enrollment, or direct APK download
• Web: Hosted at app.happier.dev or self-hosted
• Desktop: Available as part of the application suite

Self-hosting the relay server

For complete data control:

happier relay host install --mode system

This installs a system service using SQLite by default, with release channels (stable, preview, dev) selectable via --channel. Docker and Proxmox deployment options are also documented.

Real Code Examples

The README provides clear command substitutions that demonstrate Happier's drop-in replacement pattern for existing AI CLI tools.

Basic provider substitution:

# Instead of: claude
# Use: happier
happier

# Instead of: codex
# Use: happier codex
happier codex

# Instead of: opencode
# Use: happier opencode
happier opencode

These aren't aliases—Happier wraps the underlying providers while adding its cross-device session management, encryption, and continuity layer. The happier command without arguments defaults to Claude Code; explicit subcommands select other providers.

Extended provider set:

# More providers:
happier gemini
happier kilo
happier kimi
happier qwen

This demonstrates the unified interface principle: regardless of which AI agent you're using, the invocation pattern and session management remain consistent. The daemon handles provider-specific process management while presenting a common surface to clients.

Session reattachment:

# Reconnect to a running session in terminal
happier attach

This is critical for the handoff workflow—start from mobile or desktop, then reclaim terminal control without losing context, or vice versa. The session ID remains stable; you're simply changing the control surface.

Authentication with explicit method:

yarn auth login -- --method=mobile --no-open

From the development/source workflow documentation, this shows how to force mobile-first auth and suppress automatic browser opening—useful in headless or SSH environments where you want to complete authentication on your phone instead.

Advanced Usage & Best Practices

Prefer self-hosting for sensitive codebases. While Happier Cloud uses end-to-end encryption, the zero-knowledge architecture is most verifiable when you control the relay server. The Docker deployment is lightweight enough for personal infrastructure; Tailscale Serve provides mobile access without exposing ports.

Use profiles for multi-context work. Named backend configurations (endpoint, auth scheme, environment variables, secrets) let you switch between personal projects, client work, and company infrastructure without credential collision. Select at session start or via --profile from CLI.

Enable approval gating for voice actions. The voice agent's power—monitoring all sessions, answering permissions, switching context—should be matched with careful per-action approval settings. Treat it as a privileged automation, not a convenience shortcut.

Run dev builds cautiously. The dev channel (hdev command) can be installed alongside stable but is explicitly warned as "highly unstable" with partial commits. Never mix dev clients with stable servers—version skew will break sessions.

Monitor the inbox, not individual sessions. The global attention center aggregates permission requests, user-action prompts, and unread activity across all machines. For developers running multiple agents, this prevents notification fatigue and missed approvals.

Comparison with Alternatives

Happier's trade-off is complexity. Native CLIs install in seconds; Happier requires CLI + daemon + relay + client app. The payoff is meaningful only if cross-device continuity, collaboration, or enterprise control are actual pain points. For terminal-only workflows, native tools remain simpler. For teams already sharing screens or struggling with "who's watching the agent," Happier fills a specific gap that vendors haven't addressed.

FAQ

Is happier-dev/happier free? Yes, MIT licensed. You pay only for your AI provider usage (Claude, OpenAI, etc.), not for Happier itself.

Does it work without internet? Local sessions work offline, but cross-device sync requires the relay server—self-hosted on your network or cloud-hosted.

Can I use my existing Claude/Codex/OpenCode installations? Yes, Happier wraps existing provider CLIs. You don't reinstall or reconfigure them.

How secure is the encryption? TweetNaCl-based E2E encryption with zero-knowledge architecture. Keys never leave your devices; servers cannot read session content.

What's the difference from Happy? Happier is a hard fork with faster iteration, broader provider support, and additional features like voice, enterprise auth, and session forking. Both are open-source.

Is it production-ready? Alpha preview as of mid-2026. Expect bugs; the team iterates rapidly and welcomes issue reports via Discord.

Can I contribute? Yes—see CONTRIBUTING.md in the repository. The maintainers emphasize welcoming community input.

Conclusion

happier-dev/happier addresses a genuinely underserved need: AI coding agents are powerful but physically tethered. For developers who start complex agent sessions and need to step away—whether to another room, another device, or another teammate—Happier provides the continuity layer that native CLIs lack.

It's best suited for: developers running multiple AI providers who want one interface; teams needing collaborative session oversight; organizations requiring self-hosted, auditable infrastructure; and anyone who's missed a critical permission request because they weren't at their terminal.

The alpha status means you should evaluate it for your tolerance for rough edges, not mission-critical deployment. But the foundation is solid—TypeScript, MIT-licensed, Swiss-built with privacy-first architecture, and shaped by real developer feedback rather than product roadmap speculation.

Explore the repository, try the hosted cloud version, or self-host in minutes. The Discord community is active for questions and issue reports.

→ View happier-dev/happier on GitHub