diff --git a/agents/nucliadb/src/hyperforge_nucliadb/basic_ask_agent.py b/agents/nucliadb/src/hyperforge_nucliadb/basic_ask_agent.py
index ba41d79..d3e47ec 100644
--- a/agents/nucliadb/src/hyperforge_nucliadb/basic_ask_agent.py
+++ b/agents/nucliadb/src/hyperforge_nucliadb/basic_ask_agent.py
@@ -1554,6 +1554,17 @@ async def inner_facets(
             logger.info(
                 f"No labels selected for faceted search on question {question}, skipping."
             )
+            context = Context(
+                agent_id=self.agent_id,
+                original_question_uuid=memory.original_question_uuid,
+                actual_question_uuid=None,
+                missing=question,
+                question=question,
+                source=source.id,
+                chunks=[],
+                agent="basic_ask_facets",
+                title=f"Faceted search on {source.id} Knowledge Box",
+            )
         else:
             # We are assuming resources equals fields here
             facets, total = await ndb.field_labels(labelsets=labels_selected)
@@ -1593,7 +1604,11 @@ async def inner_facets(
         await memory.add_step(
             step_module=self.config.module,
             step_title=self.step_title("Faceted count"),
-            step_reason="Faceted search performed",
+            step_reason=(
+                "Faceted search performed"
+                if labels_selected
+                else "Faceted search skipped"
+            ),
             step_value=selected_labels.get("reasoning", ""),
             timeit=time() - t0,
             input_nuclia_tokens=input_nt,
diff --git a/agents/nucliadb/tests/test_facets_count.py b/agents/nucliadb/tests/test_facets_count.py
new file mode 100644
index 0000000..c2e42d7
--- /dev/null
+++ b/agents/nucliadb/tests/test_facets_count.py
@@ -0,0 +1,72 @@
+from unittest.mock import AsyncMock, Mock
+
+import pytest
+from hyperforge.models import Context, Source
+
+from hyperforge_nucliadb.basic_ask_agent import BasicAskAgent
+from hyperforge_nucliadb.basic_ask_config import BasicAskAgentConfig
+
+
+@pytest.mark.asyncio
+@pytest.mark.parametrize("label_sets", [[], ["topic"]])
+async def test_facets_count_handles_label_selection(mocker, label_sets):
+    agent = BasicAskAgent(BasicAskAgentConfig(sources=["kb"]), agent_id="facets-agent")
+    source = Source.model_construct(id="kb")
+    mocker.patch(
+        "hyperforge_nucliadb.basic_ask_agent.choose_source",
+        new=AsyncMock(return_value=[source]),
+    )
+    driver = Mock(
+        labels=AsyncMock(return_value={"topic": ["science"]}),
+        field_labels=AsyncMock(return_value=({"topic": {"science": 2}}, 2)),
+    )
+    mocker.patch(
+        "hyperforge_nucliadb.basic_ask_agent.get_ndb_driver", return_value=driver
+    )
+    memory = Mock(original_question_uuid="question-id", add_step=AsyncMock())
+    manager = Mock(
+        execute_json=AsyncMock(
+            return_value=(
+                {"label_sets": label_sets, "reasoning": "Label selection"},
+                1,
+                2,
+            )
+        ),
+        execute=AsyncMock(return_value=("There are 2 science documents.", 3, 4, 0)),
+    )
+
+    contexts = await agent.facets_count(memory, manager, question="How many?")
+
+    assert len(contexts) == 1
+    context = contexts[0]
+    assert isinstance(context, Context)
+    assert context.agent_id == "facets-agent"
+    assert context.original_question_uuid == "question-id"
+    assert context.actual_question_uuid is None
+    assert context.question == "How many?"
+    assert context.source == "kb"
+    assert context.agent == "basic_ask_facets"
+    assert context.title == "Faceted search on kb Knowledge Box"
+    memory.add_step.assert_awaited_once()
+    step = memory.add_step.call_args.kwargs
+    assert step["step_value"] == "Label selection"
+    assert step["step_agent_path"] == "/context/facets-agent/facets_count"
+
+    if label_sets:
+        assert context.missing is None
+        assert len(context.chunks) == 1
+        assert context.chunks[0].chunk_id == "facets_count_result-0"
+        assert context.chunks[0].text == "There are 2 science documents."
+        driver.field_labels.assert_awaited_once_with(labelsets=label_sets)
+        manager.execute.assert_awaited_once()
+        assert step["step_reason"] == "Faceted search performed"
+        assert step["input_nuclia_tokens"] == 4
+        assert step["output_nuclia_tokens"] == 6
+    else:
+        assert context.missing == "How many?"
+        assert context.chunks == []
+        driver.field_labels.assert_not_awaited()
+        manager.execute.assert_not_awaited()
+        assert step["step_reason"] == "Faceted search skipped"
+        assert step["input_nuclia_tokens"] == 1
+        assert step["output_nuclia_tokens"] == 2