Claude Code
on your phone.

Claude Code has no app of its own. You reach it through the Claude app, a chat app, a terminal over SSH, or approvals on your lock screen. Here is what each one needs, so you can pick.

Connect Claude Code

mcp.json

{
  "mcpServers": {
    "pushary": {
      "url": "https://pushary.com/api/mcp/mcp"
    }
  }
}

Uitgelicht op Hacker News

Claude Code, Codex, Cursor, Gemini CLI en meer

iPhone, Android en Mac

Je code verlaat nooit je machine

The ways to use Claude Code from your phone

Anthropic's own options come first. Most need Claude Code signed in through claude.ai, Channels also take a Console API key, and none of them run on Bedrock, Google Cloud or Foundry.

Remote Control in the Claude app

What it does: Drive a Claude Code session running on your computer from the Code tab in the Claude app for iOS and Android. Start it with claude --remote-control, or run /remote-control in a session that is already open. In /config you can turn on push when Claude needs a decision.
What it needs: A Pro, Max, Team or Enterprise plan, signed in through claude.ai. On Team and Enterprise an Owner turns Remote Control on in the admin settings first.
Where it stops: Not available with an API key, Amazon Bedrock, Google Cloud, Microsoft Foundry, or when ANTHROPIC_BASE_URL points at a gateway. It covers Claude Code only, and the claude process and your computer have to keep running.

Cloud sessions in the Claude app

What it does: Start a task on a GitHub repository from the Code tab. It runs on cloud infrastructure, Anthropic-managed by default, and keeps going after you put the phone away, so your laptop can be closed.
What it needs: A claude.ai account that shows the Code tab, and the repository on GitHub.
Where it stops: It runs in the cloud, so it cannot use your local files, tools or MCP servers. Like Remote Control, it is not reachable with an API key or through Bedrock.

Projects in the Claude app

What it does: A project is a conversation where Claude runs related threads of work, mostly in the cloud, and shows which ones finished or need you. A thread that needs your computer can run there through Remote Control.
What it needs: A Pro or Max plan. Threads work on GitHub repositories, or on files and Google Drive folders you upload.
Where it stops: A public beta that is still rolling out, and not offered on Team or Enterprise yet. A thread on your computer runs only while it is awake with Remote Control on.

Dispatch from the Claude app

What it does: Message a task from your phone to the Claude Desktop app on your computer. Dispatch decides how to run it, starts a Claude Code session for development work, and notifies you when that session finishes or needs your approval.
What it needs: A Pro or Max plan and the Claude Desktop app running on your computer.
Where it stops: Not available on Team or Enterprise. You hand over a task rather than steer a session you already have open, and your computer has to stay on.

Channels: Telegram, Discord or iMessage

What it does: Chat with a Claude Code session on your computer from Telegram, Discord or iMessage. The official plugins also forward permission prompts to the chat, so you can approve or deny them there.
What it needs: A claude.ai account or a Console API key, the channel plugin and Bun. On Team and Enterprise an admin enables channels first.
Where it stops: A research preview. Not available on Bedrock, Google Cloud or Foundry, and messages only arrive while the session is open.

SSH into your own computer

What it does: Open the terminal itself from an SSH app on your phone, usually over Tailscale or a VPN, with tmux keeping the Claude Code session alive between visits.
What it needs: A computer you can reach from outside, an SSH app and a little setup.
Where it stops: You get a full terminal on a small screen, and nothing tells you when Claude Code stops to ask for approval.

Pushary

What it does: Claude Code's permission prompts and questions come to your phone as push notifications, sent from its own hooks, however Claude Code signs in, Bedrock, Google Cloud and Foundry included. Approve from the notification once your phone is unlocked; deny is one tap. Codex, Cursor and Gemini CLI land in the same app.
What it needs: One setup command on your computer and the Pushary app. On the web: $9.99/mo after a 3-day free trial. Cancel anytime.
Where it stops: A separate paid service, not part of your Claude plan. Your computer has to stay on, as with Remote Control.

Which one to use

You sign in through claude.ai and only run Claude Code: start with Remote Control. It comes with your plan.
Your laptop will be closed: start a cloud session on a GitHub repository, or a project on Pro or Max.
You use the Desktop app on Pro or Max and want to hand over a whole task: message Dispatch.
You would rather talk to Claude in Telegram, Discord or iMessage, or you work on a Console API key: try Channels.
You use Bedrock, Google Cloud, Foundry or a gateway, or you run Codex, Cursor or Gemini CLI as well: that is where Pushary fits.

Updated Oct 24, 2026

Checked againstClaude Code on mobileRemote ControlChannels

You do not need your terminal on your phone

Mirroring your whole terminal to a phone screen is a worse desk. The thing you actually need on the go is the handful of decisions only you can make.

The terminal alone, no remote

Sit and watch the agent work

Catch the risky command in time, or miss it

Walk away and it stalls on a question

No record of what you approved

Nothing stops a runaway run

With Pushary on your phone

The agent runs on your machine

Risky commands wait for your tap to approve

Answer a question and it keeps going

Every decision lands in an audit trail

Kill switch stops it in one move

Approve / Deny

Risky commands wait on your machine for your answer from your phone, until your rule's wait ends.

Kill switch

One tap halts a runaway agent. No negotiating with it, no racing to the terminal.

Audit trail

Every approval, denial and auto decision is recorded. Export it on Agent Pro and above.

Let Claude Code run. Keep your hand on the controls.

One command to connect. Approve, answer, or stop from your phone.

Start your trial

Claude Code notifications·Claude Code Remote Control: limits·Human in the loop·Pushary vs Happy·Pushary vs Omnara·All agents

Frequently asked questions

Is there a Claude Code mobile app?

No. Claude Code has no separate mobile app. You use the Claude app for iOS and Android: the Code tab for cloud sessions, projects and Remote Control into a session on your computer, and Dispatch for handing a task to the Desktop app. Running /mobile in Claude Code shows a QR code that opens the right app store.

Does my computer have to stay on while I use Claude Code from my phone?

For Remote Control, Dispatch, Channels and Pushary, yes. The work runs on your computer, so it has to stay on with Claude Code or the Desktop app running; Remote Control reconnects when the machine wakes from sleep. Cloud sessions, and project threads that run in the cloud, keep going with your laptop closed.

Can I control Claude Code from my phone?

Yes. Remote Control, Dispatch and Channels are Anthropic's own ways to reach Claude Code on your computer. The Pushary app can also start Claude Code on a connected computer, send an instruction to a reachable session, approve or deny risky commands, answer questions, and stop a running agent. Your computer must be awake, logged in, online, and running the Pushary background service.

Does this put my whole terminal on my phone?

No. Pushary sends task-level instructions to the agent; it does not expose an arbitrary remote shell or mirror terminal output. Claude Code stays on your computer with its normal permissions and approval hooks. Start and Send require your phone's biometric or device passcode, and disappear when no live consumer is reachable.

How do I approve or deny commands from my phone?

Set a permission policy that says which commands need your approval. When Claude Code tries to run one, Pushary pushes it to your phone with Approve and Deny buttons, and the command waits for your answer. When the wait ends, your rule denies it, hands it to Claude Code's own prompt, or approves it.

What if I do not respond?

When the wait ends, your rule decides whether to deny the request, hand it to Claude Code's own prompt, or approve it. A wait action ignores the timeout and holds the request for your phone for up to about ten minutes, then hands it to the prompt. Silence never approves anything unless you chose auto-approve as the timeout action.

Does it work with anything besides Claude Code?

Approvals and questions work with Claude Code, Codex, Cursor, Gemini CLI, Hermes, VS Code, and MCP clients. Starting from the phone is currently supported for Claude Code, Codex, and Gemini CLI when that provider is installed and advertised by the selected machine.

How much does it cost?

On the web, Pushary starts at $9.99/month with a 3-day free trial. That includes phone approvals, the permission policy, the 100 most recent audit entries, and 5,000 notifications a month.

Claude Codeon your phone.