diff --git a/.openclaw-sync/source.json b/.openclaw-sync/source.json
index e630191c0..0ae28aac0 100644
--- a/.openclaw-sync/source.json
+++ b/.openclaw-sync/source.json
@@ -1,5 +1,5 @@
 {
   "repository": "openclaw/openclaw",
-  "sha": "e2f3044b8fb7d667edff52df8e1c7cac22cd2d08",
-  "syncedAt": "2026-04-28T10:23:15.277Z"
+  "sha": "d70191f8afd4e67cd12bf3bf8c952c7333e47f13",
+  "syncedAt": "2026-04-28T10:34:33.353Z"
 }
diff --git a/docs/gateway/sandboxing.md b/docs/gateway/sandboxing.md
index a2a3567d8..e953635e6 100644
--- a/docs/gateway/sandboxing.md
+++ b/docs/gateway/sandboxing.md
@@ -89,6 +89,8 @@ SSH-specific config lives under `agents.defaults.sandbox.ssh`. OpenShell-specifi
 
 Sandboxing is off by default. If you enable sandboxing and do not choose a backend, OpenClaw uses the Docker backend. It executes tools and sandbox browsers locally via the Docker daemon socket (`/var/run/docker.sock`). Sandbox container isolation is determined by Docker namespaces.
 
+To expose host GPUs to Docker sandboxes, set `agents.defaults.sandbox.docker.gpus` or the per-agent `agents.list[].sandbox.docker.gpus` override. The value is passed to Docker's `--gpus` flag as a separate argument, for example `"all"` or `"device=GPU-uuid"`, and requires a compatible host runtime such as NVIDIA Container Toolkit.
+
 <Warning>
 **Docker-out-of-Docker (DooD) constraints**