AISquare
rlx

Control the fleet remotely

Two ways to reach your agents away from the desk. rlx's own Remote puts the whole fleet on your phone through your ngrok tunnel; Claude Code's Remote Control, turned on per role, lets you continue one agent's session from claude.ai/code or the Claude app.

There are two different remotes, and they answer different needs:

rlx Remote (your phone)Claude Code's Remote Control
What you reachEvery project: what needs you, each agent's live screen, the boardsOne agent's Claude Code session
Where you use itA page on your phone, through your own ngrok tunnelclaude.ai/code or the Claude app
Who can get inWhoever has the secret link and the four-word passwordAnyone signed in to the Claude account the agent runs on, which a usage-limit hand-over can change
Off by default?Yes, and read-only until you allow writesYes, for every role
Where it is in rlx⚙ Settings ▸ Remote control, or RThe Spawn dialog's Claude Remote Control switch

In rlx's Settings and the R panel, Remote control means the first one, the phone remote, and the Spawn dialog calls the second Claude Remote Control. Two places use the plain name for the second: rlx doctor's remote control row and the dim line rlx launch prints.

Put the fleet on your phone

Open Remote control

In rlx, open ⚙ Settings ▸ Remote control. It opens App settings on its Remote tab.

The ⚙ Settings hub: a link for each page, App settings, Remote control, Accounts, Personas, Usage & stats, Agent resources and Doctor among them, each with a line on what it opens, Fleet's Restart all…, Stop all…, Pause all… and Resume all…, and the status line in the sidebar.

Paste your ngrok token, once

The first time, the tab asks for your ngrok token. Copy it from your ngrok dashboard (a free account has one), paste it under Paste your ngrok token and press Save. If you already signed ngrok in on this machine, it is found and there is nothing to paste.

⚙ Settings ▸ Remote control on a fresh install: the field to paste your ngrok token, once.

Turn it on and scan the QR code

Press Turn on remote control. The tab shows the link, a QR code and a four-word password. Scan the QR code with your phone, then type the password on the page that opens.

⚙ Settings ▸ Remote control, on: the link with Copy link and Turn off, the QR code and the password.

Allow writes, if you want to act

The phone starts read-only. To answer prompts and tell, stop, restart or switch agents from it, open the tab's Advanced and switch on Allow write actions. Set Auto-off there too: 30, 60 or 120 minutes, or Never.

From a shell, the same switch and its state are:

rlx remote allow-write on    # or off
rlx remote status            # on or off, the link, the password, the devices
rlx remote serve             # Remote without the fleet UI; run your own ngrok beside it

Set AISQUARE_REMOTE_NGROK_URL to your free static ngrok domain before you start rlx, so the link and the phone's sign-in survive a restart. Your fleet on your phone covers the rest: the Needs feed, notifications, agent actions, the security model, and running Remote without the fleet UI.

Continue one agent from claude.ai/code

Claude Code has its own Remote Control: a session started with it can be continued from claude.ai/code or the Claude app. rlx can start a fleet agent that way, titled with the project's folder name and the agent's label (<folder> · <label>), so you find it by name there.

Anyone signed in to that Claude account can drive the agent

A remote-controlled agent can be driven by anyone who is signed in to the Claude account it runs on, from any browser or device. It keeps its role's permission mode, so a coder in auto mode still approves routine tool calls by itself. It also keeps Remote Control when it moves: a restart and an account switch replay it, and with the default accounts.on_limit = switch a usage-limit hand-over can move it to any enabled, signed-in account in your pool, even one you named at spawn. So turn it on only when every account in the pool is one only you sign in to, or set rlx config set accounts.on_limit wait while any agent has it on. And only for the roles you mean to drive from elsewhere.

It is off for every role by default, and it needs a Claude.ai login (Pro, Max, Team or Enterprise) talking straight to the Anthropic API.

Open the Spawn dialog

In the sidebar, press + Spawn agent. Pick the Project, the Role and, if you want, the Account.

Switch on Claude Remote Control

Open Advanced and switch on Claude Remote Control ("continue it from claude.ai/code or the Claude app"). It opens at the role's default, so it is off unless you turned it on for the role.

The Spawn dialog with Advanced open: Agent, Binary, Persona, Label, Task, Worktree, Claude Remote Control switched on, Permission, Permission mode, Model set to opus and Effort set to high.

Spawn, then open it elsewhere

Press Spawn. In claude.ai/code or the Claude app, signed in to the same Claude account, the session appears as <folder> · <label>, for example api · coder-auth. Restarts and account switches keep the choice.

From a shell

For one agent, pass the flag to the spawn:

rlx fleet spawn coder --remote-control --prompt "Start on #12"
rlx fleet spawn coder --no-remote-control     # off for this one, over a role that has it on

To start every agent of a role with it, set the role's default. App settings does not show this key, so set it from a shell:

rlx config set fleet.roles.coder.remote_control true

rlx launch <role> takes --remote-control and --no-remote-control too, and otherwise follows the same role setting. rlx --json fleet ls reports each agent's choice as remote_control.

When it is not turned on

Claude Code refuses Remote Control in some setups, so the launch leaves the flag off and says why in one dim line (remote control: … — not turned on) when:

  • the role's binary is not Claude Code;
  • the launch line already carries --remote-control or --rc of your own, which wins as you wrote it;
  • ANTHROPIC_BASE_URL points at a host other than the Anthropic API, as the tracing proxy (parked) does;
  • CLAUDE_CODE_OAUTH_TOKEN is set, since that token can make model requests only.

While any role has it on, rlx doctor shows a remote control row. It names those roles, and warns when one of the blockers above would make Claude Code refuse it, reading the environment the fleet's agents actually inherit.

On this page