diff --git a/haystack/utils/base_serialization.py b/haystack/utils/base_serialization.py
index 0db98b6d39..43b30947d3 100644
--- a/haystack/utils/base_serialization.py
+++ b/haystack/utils/base_serialization.py
@@ -28,7 +28,9 @@ def _serialize_value_with_schema(payload: Any) -> dict[str, Any]:  # noqa: PLR09
     - Objects with to_dict() methods (e.g. dataclasses)
     - Objects with __dict__ attributes
     - Dictionaries
-    - Lists, tuples, and sets. Lists with mixed types are not supported.
+    - Lists, tuples, and sets, including ones holding mixed types. A mixed-type array records one
+      schema per position under the JSON Schema `prefixItems` keyword, while a homogeneous array
+      keeps the single shared `items` schema.
     - Primitive types (str, int, float, bool, None)
 
     This is for runtime values (Agent/State data, pipeline inputs/outputs at a breakpoint), not
@@ -61,20 +63,23 @@ def _serialize_value_with_schema(payload: Any) -> dict[str, Any]:  # noqa: PLR09
 
     # Handle array case - iterate through elements
     if isinstance(payload, (list, tuple, set, frozenset)):
-        # Serialize each item in the array
+        # Serialize each item in the array, keeping its own schema
         serialized_list = []
+        item_schemas: list[Any] = []
+        base_schema: dict[str, Any]
         for item in payload:
             serialized_value = _serialize_value_with_schema(item)
             serialized_list.append(serialized_value["serialized_data"])
+            item_schemas.append(serialized_value["serialization_schema"])
 
-        # Determine item type from first element (if any)
-        # NOTE: We do not support mixed-type lists
-        if payload:
-            first = next(iter(payload))
-            item_schema = _serialize_value_with_schema(first)
-            base_schema = {"type": "array", "items": item_schema["serialization_schema"]}
-        else:
+        if not item_schemas:
             base_schema = {"type": "array", "items": {}}
+        elif all(item_schema == item_schemas[0] for item_schema in item_schemas):
+            # Homogeneous array: keep the historical `items` envelope so older readers keep working
+            base_schema = {"type": "array", "items": item_schemas[0]}
+        else:
+            # Mixed-type array: describe every position with the JSON Schema `prefixItems` keyword
+            base_schema = {"type": "array", "prefixItems": item_schemas}
 
         # Add JSON Schema properties to infer sets, frozensets and tuples
         if isinstance(payload, (set, frozenset)):
@@ -194,7 +199,8 @@ def _deserialize_value_with_schema(serialized: dict[str, Any]) -> Any:
          "serialized_data": <the actual data>
       }
 
-    NOTE: For array types we only support homogeneous lists (all elements of the same type).
+    For array types, `prefixItems` (one schema per position, emitted for mixed-type arrays) takes
+    precedence over `items` (a single shared schema, emitted for homogeneous arrays).
 
     :param serialized: The serialized dict with schema and data.
     :returns: The deserialized value in its original form.
@@ -233,10 +239,23 @@ def _deserialize_value_with_schema(serialized: dict[str, Any]) -> Any:
 
     # Handle array case
     if schema_type == "array":
-        # Deserialize each item
+        # Mixed-type arrays carry one schema per position under `prefixItems`,
+        # homogeneous ones carry a single shared schema under `items`.
+        prefix_items = schema.get("prefixItems")
+        if prefix_items is not None:
+            if len(prefix_items) != len(data):
+                raise DeserializationError(
+                    f"Invalid array payload: 'prefixItems' declares {len(prefix_items)} element schemas "
+                    f"but 'serialized_data' holds {len(data)} elements."
+                )
+            item_schemas: list[Any] = list(prefix_items)
+        else:
+            item_schemas = [schema["items"]] * len(data)
+
+        # Deserialize each item with the schema of its own position
         deserialized_items = [
-            _deserialize_value_with_schema({"serialization_schema": schema["items"], "serialized_data": item})
-            for item in data
+            _deserialize_value_with_schema({"serialization_schema": item_schema, "serialized_data": item})
+            for item_schema, item in zip(item_schemas, data, strict=True)
         ]
         final_array: list | set | frozenset | tuple
         # Is a set or frozenset if uniqueItems is True
diff --git a/releasenotes/notes/fix-mixed-type-lists-schema-serialization-3f1c5a2b7d904e18.yaml b/releasenotes/notes/fix-mixed-type-lists-schema-serialization-3f1c5a2b7d904e18.yaml
new file mode 100644
index 0000000000..33827dffe1
--- /dev/null
+++ b/releasenotes/notes/fix-mixed-type-lists-schema-serialization-3f1c5a2b7d904e18.yaml
@@ -0,0 +1,9 @@
+---
+fixes:
+  - |
+    Fixed schema-based serialization of lists, tuples and sets holding mixed types. Previously the
+    schema was derived from the first element only, so deserializing such a value raised an
+    `AttributeError` or silently returned mis-typed data (for example an Agent `State` field or a
+    pipeline breakpoint input holding `[Document(...), "text", 3]`). Mixed-type arrays now record one
+    schema per position using the JSON Schema `prefixItems` keyword and round-trip correctly.
+    Homogeneous arrays keep the exact same output as before, so existing snapshots still load.
diff --git a/test/utils/test_base_serialization.py b/test/utils/test_base_serialization.py
index 67f48071c0..683f35c4a7 100644
--- a/test/utils/test_base_serialization.py
+++ b/test/utils/test_base_serialization.py
@@ -373,6 +373,86 @@ def test_serialize_and_deserialize_value_with_schema_with_various_types():
     assert _deserialize_value_with_schema(expected) == data
 
 
+class TestMixedTypeArrays:
+    """Mixed-type arrays keep one schema per position under `prefixItems`."""
+
+    def test_serialize_mixed_primitives(self):
+        assert _serialize_value_with_schema({"y": [1, "a", {"k": 2}]}) == {
+            "serialization_schema": {
+                "type": "object",
+                "properties": {
+                    "y": {
+                        "type": "array",
+                        "prefixItems": [
+                            {"type": "integer"},
+                            {"type": "string"},
+                            {"type": "object", "properties": {"k": {"type": "integer"}}},
+                        ],
+                    }
+                },
+            },
+            "serialized_data": {"y": [1, "a", {"k": 2}]},
+        }
+
+    @pytest.mark.parametrize(
+        "value",
+        [
+            pytest.param({"y": [1, "a", {"k": 2}]}, id="mixed-primitives"),
+            pytest.param({"x": [Document(content="a", id="1"), "plain string", 3]}, id="object-and-primitives"),
+            pytest.param({"t": (1, "a", Document(content="a", id="1"))}, id="mixed-tuple"),
+            pytest.param({"n": [[1, "a"], [Document(content="a", id="1"), None]]}, id="nested-mixed-lists"),
+            pytest.param({"e": []}, id="empty-list"),
+            pytest.param({"m": [1, None, True, 2.5, "s"]}, id="all-primitive-kinds"),
+        ],
+    )
+    def test_round_trip_mixed_arrays(self, value):
+        deserialized = _deserialize_value_with_schema(_serialize_value_with_schema(value))
+        assert deserialized == value
+        for key, original in value.items():
+            assert type(deserialized[key]) is type(original)
+
+    def test_round_trip_mixed_set(self):
+        value = {1, "a", None}
+        deserialized = _deserialize_value_with_schema(_serialize_value_with_schema(value))
+        assert deserialized == value
+        assert type(deserialized) is set
+
+    def test_round_trip_mixed_frozenset(self):
+        value = frozenset({1, "a", None})
+        deserialized = _deserialize_value_with_schema(_serialize_value_with_schema(value))
+        assert deserialized == value
+        assert type(deserialized) is frozenset
+
+    def test_homogeneous_output_is_unchanged(self):
+        # Backward compatibility: homogeneous arrays keep the historical `items` envelope.
+        document = Document(content="a", id="1")
+        assert _serialize_value_with_schema({"docs": [document], "nums": (1, 2)}) == {
+            "serialization_schema": {
+                "type": "object",
+                "properties": {
+                    "docs": {"type": "array", "items": {"type": "haystack.dataclasses.document.Document"}},
+                    "nums": {"type": "array", "items": {"type": "integer"}, "minItems": 2, "maxItems": 2},
+                },
+            },
+            "serialized_data": {"docs": [document.to_dict()], "nums": [1, 2]},
+        }
+
+    def test_deserialize_old_format_with_only_items(self):
+        # Payloads written before `prefixItems` existed still deserialize through `items`.
+        assert _deserialize_value_with_schema(
+            {"serialization_schema": {"type": "array", "items": {"type": "integer"}}, "serialized_data": [1, 2, 3]}
+        ) == [1, 2, 3]
+
+    def test_deserialize_prefix_items_length_mismatch_raises(self):
+        with pytest.raises(DeserializationError, match="'prefixItems' declares 2 element schemas"):
+            _deserialize_value_with_schema(
+                {
+                    "serialization_schema": {"type": "array", "prefixItems": [{"type": "integer"}, {"type": "string"}]},
+                    "serialized_data": [1],
+                }
+            )
+
+
 class TestErrorHandling:
     @pytest.mark.parametrize(
         "value",