Happy Engineering
Happy is an open-source harness for orchestrating coding agents such as Claude, Codex, Grok and Gemini across desktop, phone and remote/team servers. It aims to provide a polished, phone-first experience with one shared permission model, end-to-end encrypted mobile sync and no model lock-in.
At a Glance
- Individual software developers and AI-assisted programmers
- Engineering teams sharing coding-agent sessions
- Developers using Claude Code, Codex, Grok, Gemini or ACP-compatible agents
- Organizations wanting self-hosted or remote coding-agent infrastructure
AI Tools by Happy Engineering
(1)Happy
Multi Agent Coding App Control
Discussions
No discussions yet
Be the first to start a discussion about Happy Engineering
Latest News
Happy Agent v2026.11.5 added same-start discovery of new Gemini Antigravity models, including Gemini 4 Carbon, plus chat controls and suggested replies.
Happy Agent v2026.11.3 added Claude Fable 5.2 support, updated the Claude Code engine and introduced code-review pages.
Happy Desktop v2026.11.1-preview.5 shipped a release candidate with Happy Cloud/team features, attached-file cards and team AI-provider management.
Happy Desktop v0.0.103 added team-server setup and QR-based phone connection, Windows sandbox setup, task organization and model/session setting improvements.
Products & Services
Free MIT-licensed macOS, Windows and Linux desktop app that orchestrates Claude Code, Codex, Grok, Gemini and other agents using the user's existing provider subscriptions.
Free iPhone/iPad and Android app that pairs with a computer or remote Happy Agent, allowing users to monitor, steer and approve coding-agent work with end-to-end encrypted sessions.
Open-source agent runtime for local, remote and team servers; it provides durable sessions, workspaces, provider harnesses, sandboxing, permissions and multiplayer operation.
The original npm/terminal client that wraps Claude Code and Codex sessions and makes them controllable from the phone; the repository now recommends Desktop/Harness for multi-model teams.
Market Position
Happy positions itself as an open-source, model-agnostic alternative to single-provider agent clients and mobile remote-control tools: unlike Claude Desktop/Claude Remote Control or the ChatGPT/Codex desktop app, it combines multiple native model harnesses, phone access, durable sessions, team servers, self-hosting and an MIT license in one system. Its closest open-source alternatives are other Claude Code/Codex phone clients and terminal wrappers, while its differentiation is multi-model orchestration plus encrypted, phone-first control.
Leadership
Founders
Steve Korshakov
Built Telegram S, the open-source Actor messaging platform and Openland (YC W18); also created llama-coder, the Tact language for TON smart contracts and Supervoice. He later designed encryption at Bee, the AI wearable acquired by Amazon in 2025.
Kirill Dubovitskiy
Previously built support and finance assistants and agent skills at Robinhood. Earlier, his Facebook feed work reportedly lifted iOS ad revenue 3%; he also created the Diagrams: Show Me ChatGPT plugin, which drew 3 million diagrams and generated $9,000 per month.
Executive Team
Steve Korshakov
Co-founder / engineer
Founder profile identifies him as the builder of Telegram S, Actor, Openland, llama-coder, Tact and Supervoice, and a former encryption designer at Bee.
Kirill Dubovitskiy
Co-founder / product engineer
Previously built Robinhood assistants and agent skills, worked on Facebook feed ranking and created the Diagrams: Show Me plugin.
Founding Story
The project began as a weekend experiment after the founders realized that coding agents could work unattended but users were still stuck at their desks to steer them. They built a phone client so users could leave the desk without interrupting Claude Code, emphasizing end-to-end privacy, a free/open-source app and hands-free voice interaction.
Business Model
Revenue Model
The core app, CLI and server are free and MIT licensed; model access uses the customer's existing Claude, Codex, Grok or other provider subscriptions/keys. Happy currently monetizes voice in the mobile app through a subscription, while users may bring their own ElevenLabs agent without a Happy voice limit.
Pricing Tiers
Open-source desktop, mobile, CLI and server; no Happy plan is required for model access.
20 voice minutes every 30 days.
The App Store lists a Plus Plus Monthly in-app purchase; the current voice documentation describes up to 5 hours per 30 days on the paid tier.
No Happy voice-minute limit when using an agent configured in the customer's ElevenLabs account.
Target Markets
- Individual software developers and AI-assisted programmers
- Engineering teams sharing coding-agent sessions
- Developers using Claude Code, Codex, Grok, Gemini or ACP-compatible agents
- Organizations wanting self-hosted or remote coding-agent infrastructure
- Remote control and approval of long-running coding-agent tasks from a phone
- Orchestrating multiple AI coding models and delegating work to subagents
- Developing software while away from a desk or from a remote server
- Collaborative coding sessions for teams
- Voice-driven agent steering and permission handling
- Meta engineers requested an internal pilot after repeatedly asking to use Happy