Claude Code Remote Control alternatives and when to use them

The short answer

Claude Code's built-in Remote Control lets you drive a session running on your computer from the Claude mobile app or claude.ai/code. If you only use Claude Code and sign in with a claude.ai subscription, it's the most polished option and you should probably start there. People look for alternatives for a few specific reasons: they run Claude Code with an API key, through Amazon Bedrock or a company LLM gateway; they also use Codex or other agents and don't want a separate app for each; or they want the relay in the middle to see only ciphertext.

Everything below about Remote Control was checked against Anthropic's Claude Code documentation on September 29, 2026. The feature changes quickly, so treat the official docs as the source of truth.

What Remote Control does well

Remote Control is a window into a local session. Claude keeps running on your machine with your filesystem, MCP servers and project configuration; the phone and browser are just another screen. You can start it with claude remote-control (server mode), claude --remote-control (an interactive session that is also reachable remotely), or /remote-control inside an existing session. The VS Code extension and the Claude desktop app support /rc as well.

Its strengths are real:

• Full sync. The conversation, including subagent progress, stays in sync across terminal, browser and phone, and you can type from any of them.
• Images and files from your phone. Photos go straight to Claude; other files are downloaded to your machine and passed in as file references.
• Automatic reconnect. If your laptop sleeps or the network drops, the session reconnects when the machine comes back.
• Parallel sessions in server mode. claude remote-control allows up to 32 concurrent sessions by default, and --spawn worktree gives each one its own git worktree.
• Official push notifications. Once enabled in /config, the Claude app can notify you when a long task finishes or when a permission prompt needs you.

Where it stops

The same documentation is clear about the limits, and these are exactly why people search for alternatives:

1. claude.ai subscription auth only. Pro, Max, Team and Enterprise plans are supported; API keys are not. On Team and Enterprise, an Owner has to switch it on first.
2. api.anthropic.com only. It is unavailable on Amazon Bedrock, Google Cloud, Microsoft Foundry, or when ANTHROPIC_BASE_URL points at an LLM gateway or proxy.
3. The session must be a Remote Control process. Close the terminal or quit the app and the session goes offline. Outside server mode, one interactive process carries one remote session at a time.
4. Claude Code only. If you also run Codex or another agent, you need a separate solution for each.
5. TLS through Anthropic's API. The docs say your machine only makes outbound HTTPS requests, never opens inbound ports, and uses multiple short-lived credentials. That describes transport security; the docs don't describe it as end-to-end encrypted. For most people that's fine, but it matters if your requirement is that the relay can't read your content.

The alternatives

Anthropic's other options: Channels and Dispatch

The docs list other ways to work away from the terminal. Channels push events from Telegram, Discord or your own server into a running Claude Code session through a plugin, which suits event-driven work like "look at this CI failure." Dispatch lets you message a task from the Claude app to a paired desktop app. Both still live inside Anthropic's account system, but if you just need a way to get messages in, check them first.

SSH plus tmux

The oldest trick works for every agent and every auth method: run the agent inside tmux, SSH in from your phone, attach. Your data stays on your own SSH connection. The costs are that you have to make your computer reachable (port forwarding, a VPN or a tunnel), driving a terminal on a phone keyboard is tiring, and there are no notifications, so a paused agent waits until you happen to look.

Hook-based push tools

Several community tools use Claude Code hooks to turn permission prompts into push notifications through services like ntfy or Telegram, so you can answer with a tap. They're lightweight and easy to modify, and they're a good fit if the only problem you have is approvals blocking progress. Before picking one, check whether the third-party service can read your commands and file paths, and whether a timeout or error ends up as allow or deny.

An always-on app on your computer: BotBus

BotBus is an always-on app (a menu bar app on the Mac, a background command-line service on Linux) that discovers the coding agents on your computer, including Claude Code and Codex, and syncs their tasks to iPhone and Apple Watch. It works differently from Remote Control:

• You keep launching Claude Code the way you do now. Install the hooks from BotBus settings and keep using claude in your terminal; session status and permission prompts reach your phone. The installer only merges and removes hooks tagged as BotBus, so your own settings stay intact.
• It uses your local claude. When you start or continue a task from the phone, BotBus runs the claude you already have installed and configured, in non-interactive mode. Authentication comes from your own setup; BotBus doesn't check it. That turn is owned by BotBus, not by a terminal window.
• More than one agent. Codex, Hermes, Pi, OpenClaw and DeepSeek Harness show up in the same list, and other agents that implement ACP can connect too. Tasks are grouped by computer and project, and several computers can pair with the same phone.
• End-to-end encrypted content. Tasks, conversations, commands, approval details and notification text are encrypted on your devices. The key reaches your phone only through the pairing QR code on your computer's screen, so the relay sees ciphertext. It still sees routing metadata such as device IDs, online status, update times and data sizes, and dev previews are outside end-to-end encryption. Details are on the security page.
• Timeouts never approve. A desktop session waits up to about 30 minutes for your answer; on timeout or if BotBus quits, the decision goes back to Claude Code's own permission prompt. In a turn started from the phone, timeouts and interrupts count as a denial.

The trade-offs, stated plainly: the desktop side covers macOS (26 or later) and Linux, and the Linux version is a command-line program that doesn't yet support the Screen view (seeing and controlling the computer's screen from your phone) or dev previews. On mobile there is only the iPhone and Apple Watch app (now on the App Store); Windows and Android versions are planned. Approving, continuing, starting and interrupting tasks require that computer to be online and depend on the connector and task state. Interrupt only works for processes BotBus started itself, so sessions you opened in a terminal have to be stopped there. And unlike Remote Control, you don't get one conversation mirrored across terminal, browser and phone; the phone shows a task list and loads conversation history on demand.

Which one should you pick?

• Claude Code only, on a Pro or Max plan: use Remote Control first. It's the most complete experience.
• API key, Bedrock or a corporate gateway: Remote Control won't work, so look at SSH + tmux, a push tool, or BotBus.
• Claude Code plus Codex or other agents: a tool that lists every task in one place, like BotBus, saves switching apps.
• You only want approvals to stop blocking: a push tool is the lightest option, once you know whose servers it uses and what happens on timeout.
• You care what the relay can read: choose an end-to-end encrypted tool or your own SSH connection.

To try BotBus, download it for your computer. If approvals are your main concern, read how to approve Claude Code permission prompts from your phone.

FAQ

Can I use Remote Control and BotBus at the same time?

Yes, and they don't interfere with each other. Remote Control lives inside the Claude Code process, while BotBus works through hooks and processes it starts itself, so each uses its own channel. You could keep using Remote Control for a Claude Code conversation while BotBus shows your Codex and other agent tasks on your watch.

I run Claude Code with an API key. Can I approve from my phone?

Not with Remote Control, which requires claude.ai subscription auth. BotBus runs your local claude and doesn't check how it signs in; as long as claude works on your computer, its sessions can sync to your phone.

Is Remote Control end-to-end encrypted?

Anthropic's docs say all traffic goes through the Anthropic API over TLS with short-lived, single-purpose credentials. They don't describe it as end-to-end encrypted. BotBus encrypts content end to end, though its relay still sees the metadata needed for routing.

What happens if I close the terminal?

A Remote Control session goes offline with its process until you bring it back. Tasks you start or continue from the phone with BotBus run in processes BotBus owns, so they don't depend on a terminal window, but the computer and BotBus need to stay running and online.