fleet upgrade

Name: Fleet
Author: Fleet

Self-updates the fleet binary to the latest release published at fleetctl.ai, verifies the download against its SHA256 checksum, and replaces the binary in place. Supported platforms are darwin/amd64, darwin/arm64, linux/amd64, and linux/arm64.

Installing a new binary does not, by itself, change the code your running daemons execute — brain and the watchers keep running whatever build they were started with until restarted. Pass --restart-daemons to apply the new binary in one step: upgrade then stop and restart brain plus every running watcher (repo and org). Your agents keep going — tmux sessions survive the restart, and the new watcher resumes from label and fabric state on startup. This single command replaces the previous multi-step stop/rebuild/start dance.

Use --check to compare your current version against the latest without changing anything, --dry-run to see exactly what would be downloaded and restarted, and --version to target a specific release. After upgrading, run fleet skills install to re-sync the bundled skill library.

Flags

--check	Print current vs. latest version and exit without making changes
--dry-run	Report what would be downloaded, replaced, and restarted without writing anything
--version <X.Y.Z>	Target a specific release instead of the latest
--restart-daemons	After installing, stop and restart all running daemons (brain, repo watchers, org watchers). Tmux sessions survive

Examples

Check whether a newer release is available

fleet upgrade --check

Upgrade and apply it to running daemons in one command

fleet upgrade --restart-daemons

Preview the upgrade without writing anything

fleet upgrade --dry-run --restart-daemons

Apply a binary that auto-deploy already refreshed

fleet upgrade --restart-daemons
# Already up to date.
# Binary already current — restarting daemons to apply merged fixes.

Frequently asked questions

Does upgrading restart my running agents?

No. `--restart-daemons` restarts the daemons (brain and the watchers), not your agents. Agents run as tmux sessions that survive the restart and keep executing their current task uninterrupted; the new watcher picks up from label and fabric state on startup.

Why did --restart-daemons restart daemons even though the binary was already current?

That is the intended behavior. When an auto-deploy step has already refreshed the on-disk binary, the daemons are still running the old build until restarted — so `--restart-daemons` restarts them even when no download is needed, printing "Binary already current — restarting daemons to apply merged fixes."

What happens if a daemon fails to restart?

fleet upgrade reports how many daemons failed to restart and exits non-zero, so automation and operators see the failure instead of a silent partial restart that could leave the reactive chain down.

Do I need to re-sync skills after upgrading?

Yes. Run `fleet skills install` after an upgrade to sync the skill library bundled with the new binary to ~/.claude/skills/fleet/.