@slayzone/cli 0.36.0-beta.10 → 0.36.0-beta.12

package/README.md

@@ -18,6 +18,29 @@ systemd --user on Linux), so it restarts if it crashes and starts again when you
 log in. Names are unique per machine — `create` fails if one already exists. To
 run a hub in the foreground instead, use `npx @slayzone/hub`.
 
+## Runners on this machine
+
+A runner is an execution node: it dials out to a hub and runs terminals, agents
+and git there. Mint a join token on the hub, then:
+
+    slay runner create <name> --token <szjt1…>   # install it here and keep it running
+    slay runner ls              # every runner installed on this machine
+    slay runner logs <name>
+    slay runner stop <name>     # stop it, keep it registered
+    slay runner start <name>    # bring a stopped runner back
+    slay runner rm <name>       # stop it and remove its registration
+
+The token embeds the hub URL and its cert fingerprint, so `--token` is the only
+configuration needed. It is stored in `<root>/config.json` (owner-only) — never in
+the service unit. `--root <dir>` picks where the runner's config, credentials and
+logs live (default: the current directory), and `--allow <dir>` (repeatable) sets
+the filesystem roots it may touch. One runner per root.
+
+`create` waits until the runner has actually reached its hub, and unregisters
+itself again if it cannot — so a bad token fails loudly instead of leaving a
+service that retries forever. `rm` leaves the runner's data on disk and does NOT
+revoke it on the hub; revoke it there if the machine is going away.
+
 ## Targeting a hub
 
     slay hub use <name|url>     # point this CLI at a hub