diff --git a/src/google/adk/tools/_function_tool_declarations.py b/src/google/adk/tools/_function_tool_declarations.py
index a835cd899e..5b7c10947b 100644
--- a/src/google/adk/tools/_function_tool_declarations.py
+++ b/src/google/adk/tools/_function_tool_declarations.py
@@ -64,8 +64,12 @@ def _get_function_fields(
   # Get type hints with forward reference resolution
   try:
     type_hints = get_type_hints(func)
-  except TypeError:
-    # Can happen with mock objects or complex annotations
+  except (TypeError, NameError, AttributeError):
+    # TypeError can happen with mock objects or complex annotations. NameError
+    # / AttributeError happen when an annotation is an unresolvable forward
+    # reference at runtime, e.g. a type imported only under `if TYPE_CHECKING:`
+    # (common for the context parameter, which is excluded from the schema
+    # anyway). Fall back to the raw per-parameter annotations below.
     type_hints = {}
 
   for name, param in sig.parameters.items():
@@ -145,7 +149,10 @@ def _build_response_json_schema(
     try:
       type_hints = get_type_hints(func)
       return_annotation = type_hints.get('return', return_annotation)
-    except TypeError:
+    except (TypeError, NameError, AttributeError):
+      # An unresolvable parameter annotation (e.g. a `TYPE_CHECKING`-only
+      # import) must not block resolving the return type; fall back to the
+      # raw return annotation, which pydantic resolves below.
       pass
 
   # Handle AsyncGenerator and Generator return types (streaming tools)
diff --git a/tests/unittests/tools/test_function_tool_with_import_annotations.py b/tests/unittests/tools/test_function_tool_with_import_annotations.py
index 0d171628d3..71e8d8ea61 100644
--- a/tests/unittests/tools/test_function_tool_with_import_annotations.py
+++ b/tests/unittests/tools/test_function_tool_with_import_annotations.py
@@ -16,6 +16,7 @@
 
 from typing import Any
 from typing import Dict
+from typing import TYPE_CHECKING
 
 from google.adk.tools import _automatic_function_calling_util
 from google.adk.tools.function_tool import FunctionTool
@@ -23,6 +24,9 @@
 from google.genai import types
 import pydantic
 
+if TYPE_CHECKING:
+  from google.adk.tools.tool_context import ToolContext
+
 
 def test_string_annotation_none_return_vertex():
   """Test function with string annotation 'None' return for VERTEX_AI."""
@@ -229,3 +233,25 @@ def function_with_list(items: list[ItemModel]) -> int:
   assert processed_args['items'][0].name == 'Burger'
   assert processed_args['items'][0].quantity == 10
   assert processed_args['items'][1].quantity == 5
+
+
+def test_type_checking_only_context_annotation_builds_declaration():
+  """A context parameter annotated with a TYPE_CHECKING-only import must not
+  break declaration building for the remaining parameters."""
+
+  def test_function(x: int, tool_context: ToolContext) -> str:
+    """A tool with a context parameter.
+
+    Args:
+      x: a number.
+    """
+    return str(x)
+
+  declaration = FunctionTool(test_function)._get_declaration()
+
+  assert declaration.name == 'test_function'
+  schema = declaration.parameters_json_schema
+  # The real parameter is described, the context parameter is excluded.
+  assert 'x' in schema['properties']
+  assert 'tool_context' not in schema['properties']
+  assert schema['required'] == ['x']