Fleet Atlas
A living map of every machine. Each computer is its own territory, its sessions drifting inside — green when it's humming, amber when one needs you, red when something's blocked.
Claude Code + Codex · every machine · iPhone + Apple Watch
Monitor Claude Code from your phone.
Fleet Command: Agents is a free iPhone and Apple Watch app that watches your Claude Code and Codex sessions while you're away from the terminal. Get a push notification the second a session needs you, approve or deny a command remotely from the lock screen, and answer an agent's question without going back to your desk.
Your vendor's app shows you one session. This is the layer above: every machine you own on one live space-map — who's working, who's waiting, and which laptop has gone dark. Machine heartbeats catch a dead box, not just a stalled session.
Free · works alongside Remote Control, Conductor & claude-squad · not affiliated with Anthropic or OpenAI
FLEET COMMAND
5 SHIPS · 1 WAITING
payments-api
search-svc
mobile-app
infra-terraform
docs-site
PENDING APPROVAL · mobile-app
Run a command?
$ git push --force-with-lease origin main
Deny
Approve
∞
machines per fleet
1
command to link
~5s
wrist glance
0
code or prompts sent
// HOW IT WORKS
Monitor Claude Code remotely in three steps
1
Install the plugin
On any machine running Claude Code (Mac, Windows, or Linux), add the marketplace and install — then restart Claude Code.
$ claude plugin marketplace add \ sideffect263/fleet-commander $ claude plugin install fleet-commander
2
Add a fleet
Open the app, tap Add a Fleet, name your machine. It shows a one-time pairing code.
pairing code
FLEET-7Q2K
3
Watch ships stream in
Link with the code. Every session becomes a ship. Optionally gate risky commands on phone approval.
> /fleet-link FLEET-7Q2K ✓ linked · fleet is live > /fleet-approvals on
// WHY IT'S DIFFERENT
Built for the glance, not the dashboard
The Muster
One ranked list of exactly who needs you across the whole fleet. Only what matters sends a push notification — quiet digests roll up the rest, so your phone never floods.
Approve from anywhere
Gate risky commands. Approve or deny a force-push right from the lock screen or your Apple Watch — and a fan-out of identical requests clears in one tap.
Reply to your agents
Answer an agent's question by typing from your phone — or dictating from your wrist — straight into the running session. No trip back to the terminal.
Fleet Insights
What your agents are doing, token velocity, busiest hours, and which tools and skills every session uses — across the fleet or per machine.
One fleet, every machine
Mac, Windows, or Linux; Claude Code & Codex in the same fleet. Machine heartbeats tell the two failures apart: a session stuck mid-work on a live machine shows Adrift; a laptop that stops beating shows Offline — you know when the machine itself went dark, not just the session.
// SEE IT LIVE
Seventeen seconds of fleet
Straight from the app: the living map, an approval cleared from the ship's side, and Insights on what your fleet is doing.
// INSIDE THE APP
What monitoring Claude Code looks like
Eight screens from the iPhone and Apple Watch app — the live fleet map, the notification that reaches you first, and the approval you clear with one tap.
9:41
5 SHIPS
mobile-app
// THE FORM FACTOR OTHERS LACK
Claude Code on your Apple Watch
Agents run for minutes. You shouldn't have to babysit a terminal. Raise your wrist: green ships are thinking, a pulse means one is waiting on you. Tap to see the ship, approve the command, and put your arm back down — see what the watch app does.
- Working — agent is heads-down
- Waiting — needs your input or approval
- Idle / done — ready for the next task
// A LAYER, NOT A LOCK-IN
Plays well with everything you already run
Works alongside
Remote Control, Conductor, claude-squad — keep them. Fleet Command watches through the CLI's own hooks, so hooks fire no matter who launched the session. Everything still rolls up into one fleet.
Nothing in between
We watch the official CLI through its own hooks. Nothing sits between you and Anthropic. Nothing that can be banned, no wrapper to break on the next release.
Metadata only
By architecture, no transcripts, no code files, no prompts ever leave your machine. The only prose that ever transits is what you explicitly opt into. Read exactly what's sent →
How it compares to Remote Control, Happy & Omnara — honestly →
Bring your fleet home
Download the app, install the free plugin, and command your agents from anywhere.