Taskhold vs Claude Code Remote Control
By Jake Marsh, who builds Taskhold. Updated 2026-10-31. Facts about other tools checked 2026-10-31. 6 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 Something out of date?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: Remote Control connects the Claude app or claude.ai/code to a Claude Code session on your machine, so you can type, answer permission prompts and read diffs from your phone. Taskhold doesn't touch the session. It is a task list with a live board that Claude Code and other MCP agents report steps, progress and questions to. Many people use both.
Remote Control is part of Claude Code. You start it with claude remote-control, claude --remote-control or /remote-control, and the session shows up in the Claude app on iOS and Android and at claude.ai/code. Claude keeps running on your machine; the phone is a window into that session, with the conversation, a diff pane and push notifications.
Taskhold sits one level up. Your agents take tasks from a shared list, plan them into steps and report progress to a live board you open in any browser. It works with Claude Code and every other MCP client, and it can't type into a session or approve a tool. If you mainly want to know where your agents are, read how to see what your agent is doing; for answering questions away from your desk, see answering a blocked agent from your phone. The phone app Happy is another way to drive sessions remotely.
Side by side
| Claude Code Remote Control | Taskhold | |
|---|---|---|
| Price | Included with Claude Pro, Max, Ultra, Team and Enterprise plans; API-key logins aren't supported | $12/month or $120/year per person, 7-day trial |
| What you control from the phone | The session itself: send messages and files, answer permission prompts and questions, switch model or effort | The task list: add and order tasks, reply to an agent's question, unblock a task |
| Agents | Claude Code (CLI, Desktop app or VS Code extension) | Any MCP client: Claude Code, Codex, Cursor, Windsurf, VS Code, Gemini CLI and more |
| Several sessions | One remote session per interactive process; server mode (claude remote-control) serves up to 32 concurrent sessions by default | Every agent on every machine reports to one board, filtered by project |
| Watching progress | The conversation as it happens, plus running subagents and workflows | Steps with a bar, percent or N of M, the agent's own message, time elapsed and an estimate |
| Permission prompts | Forwarded to the device and held open until you answer | Not handled; approve them in the agent or set allow rules |
| Notifications | Push to the Claude app for permission prompts and questions, or when Claude decides to tell you | Browser notifications while the board is open; ntfy and Pushover scripts for phone pushes |
| Sharing | Not covered in the Remote Control docs | Read-only share link for a project's live board |
| After the session ends | The session goes offline when the claude process stops; server sessions can be resumed for about four hours | Each task keeps its steps, notes, summary and history, with every change marked human or agent |
Remote Control facts from code.claude.com/docs/en/remote-control, checked 2026-10-31.
What Remote Control does well
Remote Control is the full session on another screen. Your local MCP servers, files and project settings stay available, typing @ autocompletes paths from your project, and photos you attach from the phone reach Claude directly. Permission prompts and AskUserQuestion questions stay open until you answer them from any connected device, and with Push when actions required turned on in /config, the Claude app pings you when one appears. If the laptop sleeps or the network drops, Claude Code reconnects when the machine is back online.
The limits come from the same design. The claude process has to keep running on a machine that stays on, the sessions are Claude Code sessions only, and on Team and Enterprise plans an Owner has to turn Remote Control on first.
What Taskhold adds, and what it can't do
- One board across agents and machines: a Claude Code session, a Codex run on another laptop and a Cursor agent all show up as tasks with steps, so you see how far each has got without opening any of them.
- Steps and progress the agent reports on purpose (taskhold_plan, taskhold_progress), with a status dot that turns Quiet and then No word when an agent stops reporting.
- Questions about the work: the agent calls taskhold_block with one question, you reply from a phone browser, and the agent reads it with taskhold_updates. The question waits on the task even if the session has ended.
- A read-only share link, so someone else can watch a project's board without an account.
- History: what was done, by which agent, with the summary it left, after the session is long gone.
Taskhold can't drive a session. You can't type a prompt into Claude Code from the board, and it doesn't see or approve permission prompts; those stay in Claude Code, in Remote Control or in your allow rules. It doesn't start agents either. The MCP tools reference lists what agents can call.
When is Remote Control the better choice?
- You run Claude Code and want to keep steering one session from the couch or another computer.
- You need to approve permission prompts or answer Claude's questions away from the terminal.
- You already pay for a Claude plan and want nothing extra to set up.
When is Taskhold the better choice?
- Your agents are spread across Claude Code, Codex, Cursor or more than one machine.
- You want a backlog the agents pull from, with progress per step, instead of a conversation to scroll.
- You want to check on work, or share it, without opening a session.
Using both
Start Claude Code with claude --remote-control and have it work from Taskhold: it takes the next task, plans the steps and reports progress on the board. Approve permission prompts in the Claude app when it pings you, and use the live board to see where every task stands, including the ones running in other agents. Adding mcp__taskhold__* to your allow rules keeps Taskhold's own calls from showing up as prompts. Setup is on the Claude Code page.
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
- Does Remote Control work with Codex or Cursor?
- No. Remote Control is a Claude Code feature, started from the CLI, the Claude Desktop app or the VS Code extension. Taskhold works with any MCP client, so Codex and Cursor tasks show on the same board.
- Can Taskhold approve a permission prompt?
- No. Permission prompts stay in the agent. Answer them in the terminal, in the Claude app through Remote Control, or avoid them with allow rules for the tools you trust.
- Does either one keep working when my laptop is closed?
- The agent itself stops in both cases, because it runs on your machine. Taskhold's board is hosted, so you can still read the steps, notes and questions the agent left, and reply. The agent reads your reply when it runs again.
Sources
- Claude Code docs: Remote Control, checked 2026-10-31
Related
Updated 2026-10-31. Something wrong or out of date? Email hello@greatwork.company.