Local execution · Remote command · No compromises

Run your fleet.
From anywhere.

The operating system for people running real AI agents.

Local execution where the work happens. Remote command from anywhere. One system. Two surfaces.

PRODUCT

One control plane. Local execution.

FleetCrown is built for operators already running multiple AI agents across multiple projects. It makes the work visible, steerable, and recoverable.

Control

See the whole fleet at once.

Every project, active agent, queue item, and handoff sits in one operating view instead of disappearing into terminal tabs.

Live sessions · autonomy levels · dispatch

Runner

Execution stays on your machine.

The local runner owns Zellij, git, agent launching, and handoff files. FleetCrown coordinates the work without turning your environment into a cloud sandbox.

Desktop app · daemon fallback · agent tokens

Beacon

Human judgment appears at the right moments.

Queue, Beacon, Continuous, and Mission modes let each project ask for oversight only when the next decision actually needs you.

Approvals · voice intent · remote command

EXECUTION

The work happens locally.

Your agents run on your machines with full access to your environment, tools, and context. No remote sandbox limitations.

CONTROL

Command from anywhere.

The web portal gives you complete visibility and control over your entire fleet — whether your laptop is open or not.

LIVE SURFACE

Built around the state real agent work produces.

The product is not a chat box. It is an operational surface for sessions, queues, handoffs, and machine-local execution.

projects/mainContinuous

❯ migrate auth from sessions to jwt

read_file src/proxy.ts · inspect callbackUrl · patch matcher

active projects

agents running

queued intents

HOW IT WORKS

One system. Two surfaces.

Install the local runner

A native application on your machines that actually executes agents in your terminal environment (Zellij, Claude, Grok, Codex, etc.).Download →

Control from the web

The portal gives you fleet overview, per-project autonomy controls, queues, handoffs, and the ability to steer agents from anywhere.

One source of truth

Your local machines do the work. The web orchestrates. Both surfaces reflect the same reality.

DESKTOP APP

Get Fleet Runner

FleetCrown runs in your browser as a full control plane. Fleet Runner is the optional desktop app that lets agents act on your computer — open files, run commands, drive terminal sessions — while you stay in command from the web or your phone.

Web (fleetcrown.com)

Already available — no install

Full fleet visibility across all projects
Browse history, projects, and queues
Dispatch commands from any browser or phone
Read-only without a connected local runner

Desktop (Fleet Runner)

Adds local execution

Actually runs agents on your machine
Drives Zellij sessions and handoffs
Native notifications when an agent finishes
Keeps working after you close your browser

You can start with the web today and add Fleet Runner whenever you want agents to actually do work on your machine.

Download .deb (Ubuntu / Debian / Mint)(Recommended · ~80 MB · installs via package manager)

AppImage (other distros)

Open a terminal and paste this one line. It installs Fleet Runner system-wide, then launches it. No file-manager dance, no KDE security popup:

sudo dpkg -i ~/Downloads/Fleet-Runner-linux-amd64.deb && fleet-runner

After install

Make it runnable, then open

On Linux, downloads start non-executable for safety. Paste the one-line command shown under Download to mark Fleet Runner executable and launch it. On macOS and Windows (coming soon), a normal double-click is enough.

Use the same FleetCrown account you signed up with on the web. The desktop app opens straight to your dashboard. From the web, you can also click "Open in Fleet Runner" to log the desktop app in without copy-pasting a token. From v0.3.0 onward, Fleet Runner checks for updates on launch and downloads them in the background — you'll never have to manually re-download.

Dispatch your first intent

Pick a project on your computer and dispatch an intent. Fleet Runner launches the agent in a terminal session and pings you when it hands off — even if the app is hidden. The terminal session manager ships inside Fleet Runner; you just need at least one agent CLI installed (Claude Code or Grok — see prerequisites below).

What Fleet Runner uses on your computer

Fleet Runner doesn't replace the tools you already use — it drives them. Only one thing must exist on your machine for an agent to actually run: a supported agent CLI. Pick whichever AI you prefer; you only need one to start. (Zellij — the terminal session manager that gives each agent its own pane — ships inside Fleet Runner since v0.2.0, so you no longer install it separately.)

Claude Code

Anthropic's coding agent

Required

One supported agent CLI is required — Claude Code is the recommended default. Skip this if you already plan to install Grok below; you only need one.

npm install -g @anthropic-ai/claude-codeInstall Claude Code

Grok Build

xAI's coding agent

Optional

Alternative agent CLI. Install instead of Claude Code, or alongside it if you want both available to the fleet.

curl -fsSL https://x.ai/cli/install.sh | bashInstall Grok CLI

Zellij

Bundled — no install needed

Optional

Fleet Runner v0.2.0+ ships with a known-good Zellij inside the bundle and prefers it over any system install. Listed here only so you know what's running. Install it yourself only if you also want a system-wide Zellij outside Fleet Runner.

Zellij docs (optional)

GET STARTED

Choose your entry point.

Run locally

Install or build the Fleet Runner, connect an agent token, and let your real machine become the authoritative execution surface.

Download runner →

Open the control plane

Create an account, add projects, and command agent sessions from the hosted web surface.

Start building →

Read the architecture

Understand the local-runner, remote-control, queue, and handoff design before committing your workflow to it.

View whitepaper →

THE DIFFERENCE

Not another coding agent.

Most tools help you write code faster in one file or one project. We help serious builders run and orchestrate real agent operations at fleet scale.

Local execution is the foundation

Your agents run on your machines with full access to your environment. We do not force everything through remote sandboxes.

Fleet orchestration, not single-agent assistance

Built for people already running many agents across many projects. Explicit per-project autonomy levels instead of one generic agent.

Local runner + web command center

The desktop application executes. The web portal gives you fleet visibility and remote control. Two surfaces, one system.

Begin

For builders running real agent operations.

Run your fleet.From anywhere.