diff --git a/content/automation/cli.md b/content/automation/cli.md
index da2ee59..5b5bace 100644
--- a/content/automation/cli.md
+++ b/content/automation/cli.md
@@ -58,7 +58,7 @@ full list of namespaces and their actions.
 | `location` | — | `list`, `get` |
 | `project` | — | `create`, `list`, `get`, `update`, `delete`, `usage` |
 | `role` | — | `create`, `list`, `get`, `delete`, `grant`, `revoke`, `users`, `bind` |
-| `deployment` | `deploy`, `d` | `list`, `get`, `deploy`, `delete`, `revisions`, `pause`, `resume`, `rollback`, `metrics`, `set image` |
+| `deployment` | `deploy`, `d` | `list`, `get`, `deploy`, `delete`, `revisions`, `pause`, `resume`, `restart`, `rollback`, `metrics`, `set image` |
 | `domain` | — | `create`, `get`, `list`, `delete`, `purgecache` |
 | `route` | — | `create`, `get`, `list`, `delete` |
 | `waf` | — | `get`, `list`, `set`, `delete`, `metrics`, `limitmetrics` |
@@ -108,6 +108,10 @@ deploys deployment set image web \
   --project acme --location gke.cluster-rcf2 \
   --image registry.deploys.app/acme/web@sha256:…
 
+# restart a deployment (recreate its pods, same config)
+deploys deployment restart \
+  --project acme --location gke.cluster-rcf2 --name web
+
 # roll back to a previous revision
 deploys deployment rollback \
   --project acme --location gke.cluster-rcf2 \
diff --git a/content/automation/mcp.md b/content/automation/mcp.md
index 907fff1..5eed749 100644
--- a/content/automation/mcp.md
+++ b/content/automation/mcp.md
@@ -128,6 +128,7 @@ Ask in natural language; the assistant maps it to an action:
 | "what's the usage report for `acme` this month?" | `billing.report` |
 | "add the domain `app.acme.com` to `acme`" | `domain.create` |
 | "pause the `worker` deployment in `acme`" | `deployment.pause` |
+| "restart `web` in `acme`" | `deployment.restart` |
 
 A typical exchange:
 
diff --git a/content/deployments/overview.md b/content/deployments/overview.md
index 8ddc6c0..d4db4a2 100644
--- a/content/deployments/overview.md
+++ b/content/deployments/overview.md
@@ -38,6 +38,11 @@ Two actions hang off the top right:
 - **Deploy New Revision** — opens the deploy form pre-filled with the current
   configuration. Submit it and the new revision rolls out alongside the old one
   with health-checked traffic shifting.
+- **Restart** — re-rolls the current revision so the pods are recreated, without
+  changing any configuration. Useful to clear bad in-memory state or pick up an
+  external change. Available for services and workers (types that keep standing
+  pods) — not for cron jobs or static sites, which have none. Requires the
+  `deployment.deploy` permission.
 - **Pause** — stops the workload without deleting it. Resume restores it from
   the same revision. Useful for cost control on staging or for emergency stops.