Taskhold vs Happy (Happy Coder)
By Jake Marsh, who builds Taskhold. Updated 2026-10-31. Facts about other tools checked 2026-10-31. 5 min read.
We make Taskhold. Facts about other tools come from their own sites and docs, checked 2026-10-31. Where a tool doesn't describe a feature, we say so rather than saying it lacks it. Something out of date? Email hello@greatwork.company.
Short answer: Happy is a free, open-source (MIT) app that runs Claude, Codex, Grok, Gemini and other agents on your computer and lets you steer the same sessions from your phone, end to end encrypted. Taskhold doesn't run or steer agents. It is a hosted task list at $12 a month with a live board where any MCP agent reports steps, progress and questions.
The original Happy CLI wrapped Claude Code and Codex in your terminal so you could continue from your phone. It was published on npm as happy-coder, which is why many people still call it Happy Coder; the package is now happy.
Today Happy is a desktop app for macOS, Windows and Linux plus iOS and Android apps, and the original CLI (happy claude, happy codex) still works. From the phone you can send messages with images, answer permission requests, stop a turn, switch models, start sessions and read file and Git changes.
Taskhold answers a different question: which tasks your agents are on and how far each has got. Agents take work from a shared list, plan it into steps and report to a live board. For Claude Code alone, Remote Control is Anthropic's built-in way to drive a session from your phone, and the roundup compares task tools for agents.
Side by side
| Happy | Taskhold | |
|---|---|---|
| Price | Free; models run on your own Claude, Codex and Grok subscriptions. Voice is the one paid feature (20 free minutes per 30 days, then an in-app subscription) | $12/month or $120/year per person, 7-day trial |
| Open source | Yes, MIT | No |
| What it is | A harness that runs the agents on your computer, with desktop, phone and web apps | A hosted task list and live board your agents connect to over MCP |
| Agents | Claude, Codex, Grok, Gemini through Antigravity, OpenCode and Agent Client Protocol agents | Any MCP client: Claude Code, Codex, Cursor, Windsurf, VS Code, Gemini CLI and more |
| From your phone | Steer the session: messages, permission answers, model switches, new sessions, diffs | Follow tasks, reply to an agent's question, add and reorder tasks |
| Watching progress | The live session and its file and Git changes | Steps with a bar, percent or N of M, the agent's current message and time per task |
| Computer has to stay on | Yes, the agents run on it | The agents run wherever you start them; the board is hosted and readable any time |
| Privacy | End to end encrypted; Happy's relay sees ciphertext only | Tasks, steps, notes and questions are stored on Taskhold's server; agents send only what they write into those |
| Sharing | Multiplayer: invite someone into a session and either of you can steer | Read-only share link for a project's live board |
Happy facts from its GitHub README and happy.engineering docs, checked 2026-10-31.
When is Happy the better choice?
- You want to drive live agent sessions from your phone: type, approve permission requests, switch models.
- You want it free and open source, with the code on GitHub to read or fork.
- You use several model subscriptions and want one app for all of them.
- You want end to end encryption between your computer and your phone.
When is Taskhold the better choice?
- You want a backlog your agents pull from, so the next session knows what to work on without you typing it.
- You want progress per step across agents you start in different ways, including Cursor, Windsurf and VS Code agents you run outside Happy.
- You want a record of each task, its steps and the agent's summary after the session is gone.
Using both
Happy runs and steers the sessions; Taskhold holds the queue. Add Taskhold's MCP server to the agents Happy runs (Happy's docs cover skills and MCP), put the agent instructions in CLAUDE.md or AGENTS.md, and each session takes its next task from the list and reports steps to the board. You answer permission requests in Happy and check overall progress on the live board.
What about Omnara?
Omnara shows up in older comparisons of phone apps for coding agents. Its changelog says it relaunched on September 2, 2026 as the open-source alternative to Claude Managed Agents, and its README now calls it a managed agent platform: you run agents through its API on your choice of model and machine.
Omnara is Apache 2.0 licensed; self-hosting is free, and Omnara Cloud has no platform fee and is billed pay as you go, with models at provider token rates and machines by active time. That puts it closer to running agents in production than to following your own coding sessions, so it isn't a direct alternative to either Happy or Taskhold.
See how far along your coding agents are.
$12/month or $120/year after the trial. Card up front. Cancel before day 7 and you pay nothing. Every agent you run is included.
Questions
- Is Happy free?
- Yes. Happy is free and MIT licensed, and the models run on your own subscriptions. Voice in the phone app is the only paid feature: 20 free minutes every 30 days, then a subscription, or no limit from Happy with your own ElevenLabs agent.
- Is Happy Coder the same as Happy?
- Yes. happy-coder was the npm package of the original Happy CLI, which has moved to the happy package. The project lives at github.com/slopus/happy.
- Can Taskhold control an agent session like Happy does?
- No. Taskhold can't type into a session or approve permission requests. Agents report to it over MCP, and you reply to their questions on the board.
Sources
- Happy README (slopus/happy), checked 2026-10-31
- Happy docs: welcome and pricing, checked 2026-10-31
- Happy docs: phone app, checked 2026-10-31
- Happy docs: voice, checked 2026-10-31
- Omnara website, checked 2026-10-31
- Omnara README, checked 2026-10-31
- Omnara changelog, checked 2026-10-31
Related
Updated 2026-10-31. Something wrong or out of date? Email hello@greatwork.company.