diff --git a/headroom/cli/proxy.py b/headroom/cli/proxy.py
index 1ab492acc..1bb88adaf 100644
--- a/headroom/cli/proxy.py
+++ b/headroom/cli/proxy.py
@@ -1273,9 +1273,13 @@ Press Ctrl+C to stop.
 
         import asyncio as _asyncio
 
-        from headroom.memory.adapters.watchdog import EmbeddingServerWatchdog
-
         async def _start_embed_watchdog() -> Any:
+            # Import lazily inside the guarded coroutine. The sidecar module is
+            # optional and may be absent; keeping the import here lets the
+            # try/except below fall back to the per-worker embedder instead of
+            # crashing the proxy at startup with ModuleNotFoundError.
+            from headroom.memory.adapters.watchdog import EmbeddingServerWatchdog
+
             wd = EmbeddingServerWatchdog(socket_path=_embed_socket)
             await wd.start()
             ok = await wd.wait_until_healthy(timeout=30.0)
diff --git a/tests/test_cli_proxy_embedding_server.py b/tests/test_cli_proxy_embedding_server.py
new file mode 100644
index 000000000..df849a850
--- /dev/null
+++ b/tests/test_cli_proxy_embedding_server.py
@@ -0,0 +1,31 @@
+"""Regression test for `proxy --embedding-server` startup fallback.
+
+The optional embedding-server sidecar module
+(`headroom.memory.adapters.watchdog`) is not present on main, yet the
+`--embedding-server` flag advertises a graceful fallback to the per-worker
+embedder. A misplaced import made the flag raise ``ModuleNotFoundError`` at
+startup and crash the proxy instead of falling back.
+"""
+
+import sys
+
+from click.testing import CliRunner
+
+from headroom.cli import main
+
+
+def test_embedding_server_missing_sidecar_falls_back(monkeypatch):
+    # Make the optional sidecar module unimportable regardless of whether it is
+    # installed, so the fallback path is exercised deterministically.
+    monkeypatch.setitem(sys.modules, "headroom.memory.adapters.watchdog", None)
+
+    # Don't actually start a server.
+    import headroom.proxy.server as server_mod
+
+    monkeypatch.setattr(server_mod, "run_server", lambda *args, **kwargs: None)
+
+    result = CliRunner().invoke(main, ["proxy", "--embedding-server", "--port", "8799"])
+
+    assert result.exit_code == 0, f"proxy crashed instead of falling back: {result.output}"
+    assert result.exception is None
+    assert "Falling back to per-worker embedder" in result.output