Quickstart
1. Decide: web or desktop?
FleetCrown has two surfaces that share the same account and the same data:
- Web (fleetcrown.orangecat.ch) — works in any browser. Best for monitoring fleets, reviewing handoffs, and dispatching when you're not at the agent's machine.
- Desktop (Fleet Runner) — required if you want agents to actually run on this machine. Same UI as the web app, plus the local runtime that drives terminals and the OS notifications when runs finish.
You can start with web and add the desktop app whenever you want local execution. They'll merge automatically as long as you sign in with the same account.
2. Sign in
Visit /sign-in and sign in with GitHub. First time only: GitHub asks you to authorize FleetCrown. After that you land on the dashboard.
3. Install Fleet Runner (if you want local execution)
- Visit /download. The page auto-detects your OS.
- Linux: download the AppImage, then paste the one-line command on the page to mark it executable and launch. Or grab the .deb if you're on Ubuntu / Debian.
macOS: download the .dmg, drag Fleet Runner to Applications. First launch: control-click → Open (one-time Gatekeeper bypass — we aren't code-signed yet).
Windows: run the .exe. SmartScreen will warn — click “More info” → “Run anyway”. - Fleet Runner opens to the same FleetCrown interface you saw in the browser. If your browser is signed in, the desktop app is signed in automatically (it shares cookies).
- Alternatively, from /sign-in → Settings → Agent tokens, click Open in Fleet Runner to deep-link an auth token into the desktop app without copy-paste.
4. Install an agent CLI
Fleet Runner doesn't bundle the AI agent itself — it drives whatever agent CLI you install. Pick one:
- Claude Code (Anthropic) — recommended default. Install:
npm install -g @anthropic-ai/claude-code. Sign in once with your Anthropic account. - Grok CLI (xAI) — alternative. Install:
curl -fsSL https://x.ai/cli/install.sh | bash. Set yourXAI_API_KEY.
You only need one. Zellij (the terminal session manager Fleet Runner uses internally) ships inside the app since v0.2.0 — no separate install.
5. Register a project
In the dashboard, go to Projects. Add a project with a name and the absolute path to its directory on your machine. Fleet Runner uses this path to launch the agent in the right working directory.
If you have ~/.config/agent-projects.conf on disk, projects are picked up from there too — one per line, format: tab-name|directory|adapter.
6. Dispatch your first intent
Go to Control. Pick a project. Type a prompt or choose one of the built-in intents (e.g. “next_best”, “quality”, “deploy_check”). Hit dispatch.
Fleet Runner opens a Zellij session in a new terminal, launches the agent inside, and tails its session output. When the agent writes its handoff (per Claude Code's session.md convention), Fleet Runner ingests it and surfaces an OS notification: “agent idle — done: X, next: Y, health: good.”
7. Watch from anywhere
The same dashboard works from your phone (web) while a long agent runs on your laptop. Status, handoffs, and the ability to cancel a run are all live — useful when you're not at the machine.
Need help?
- Open a GitHub issue — bugs, feature requests, or questions.
- Roadmap — what's shipping next.
- Whitepaper — the bigger architecture and product thesis.