Cenvero Fleet is a self-hosted fleet management platform: shell access, service control, logs, live metrics, firewall rules, secure file transfers, live directory sync, scheduled jobs, and AI-agent control — all from one binary. Agents install themselves and connect either way, so servers behind NAT or a firewall need no inbound access. Every channel is encrypted and host-key-pinned.
brew tap cenvero/fleet && brew install cenvero-fleet
curl -fsSL https://fleet.cenvero.org/install | sh
# After install, upgrade anytime with brew upgrade cenvero-fleet
curl -fsSL https://fleet.cenvero.org/install | sh
# Also available via Homebrew — brew tap cenvero/fleet && brew install cenvero-fleet
irm https://fleet.cenvero.org/install.ps1 | iex
# Run from an elevated PowerShell prompt. Installs to %ProgramFiles%\fleet and adds it to PATH.
v2 turns Fleet into a file-moving, AI-driven control plane — all over the same authenticated SSH channel you already use.
Move files with fleet file, a dual-pane terminal manager (fleet files), or a localhost web UI (fleet file ui). Chunked, parallel, checksummed, and resumable.
Pick a local folder and a server folder with fleet sync. Fleet pushes it once, then keeps it mirrored as you edit — live, until you stop the command.
Run fleet skill claude and your AI coding agent learns the whole CLI from fleet context — then it can control any server and guide you, with your keys.
Fleet is built to be driven by Claude Code or Codex, not just humans. One command installs a tiny skill; the agent learns the rest from the binary and operates your servers — with your keys, on your machine, over SSH. No hosted control plane.
fleet skill claude # then your agent runs `fleet context` and takes it from there
Self-describing by design: fleet context and fleet ai <command> generate full, machine-readable help straight from the command tree — always in sync with the installed version, nothing to maintain by hand.
Whole deployments, not just single commands. An agent can carry a release end to end without leaving the CLI: upload the build (chunked, parallel, resumable), extract it, render config from templates and secrets, restart the service, then verify with status, logs and health checks — fanning out across the whole fleet with fleet exec --all or one tier with --group role=web. Long steps detach as jobs it can poll, and fleet guard arms a dead-man's switch that auto-reverts a risky change unless it's confirmed, so an unattended agent that locks itself out recovers on its own.
No agents to pre-install, no cloud accounts, no dashboards to subscribe to. Just a binary and SSH.
Add a server and fleet SSHes in, downloads the right agent binary, and sets up a systemd service automatically.
Direct mode for reachable servers, reverse mode for servers behind NAT or firewalls. Mix both in one fleet.
Open a root shell with fleet ssh or run commands across all servers concurrently with fleet exec --all.
Track, start, stop, and restart systemd services. Stream and cache logs for any tracked service.
Manage UFW rules and open or close ports on any server directly from the controller.
A full TUI with server health, metrics, services, logs, and alerts — all in one terminal window.
Metric threshold alerts with acknowledge and suppress actions. No external alerting system needed.
Every release is signed with minisign. The installer verifies the signature before touching your binary.
One command updates the controller and rolls out the same version to every managed agent automatically.
# 1. Install (macOS — recommended) brew tap cenvero/fleet brew install cenvero-fleet # Linux / macOS — one-liner curl -fsSL https://fleet.cenvero.org/install | sh
# 2. Initialize the controller
fleet init
# 3. Add a server (interactive) fleet server add # or non-interactive fleet server add web-01 192.0.2.10 \ --login-user root
# 4. Use it
fleet ssh web-01
fleet exec web-01 "uptime"
fleet dashboard