diff --git a/python/packages/chatkit/AGENTS.md b/python/packages/chatkit/AGENTS.md
index d11724b3ad..e99e011ecd 100644
--- a/python/packages/chatkit/AGENTS.md
+++ b/python/packages/chatkit/AGENTS.md
@@ -11,6 +11,9 @@ Integration with OpenAI ChatKit (Python) for building chat UIs.
 `ThreadItemConverter.structured_input_to_input()` is an async override hook returning
 `Message | list[Message] | None`; `to_agent_input()` normalizes its output and preserves thread order.
 
+`ThreadItemConverter.generated_image_to_input()` is an async override hook returning
+`Message | list[Message] | None`; data URI media types are inferred by `Content.from_uri()`.
+
 ## Usage
 
 ```python
diff --git a/python/packages/chatkit/README.md b/python/packages/chatkit/README.md
index ac154fbec2..94516fb06c 100644
--- a/python/packages/chatkit/README.md
+++ b/python/packages/chatkit/README.md
@@ -16,6 +16,11 @@ Override the async `ThreadItemConverter.structured_input_to_input()` method to
 customize structured input context, such as formatting or redacting answers.
 Return a `Message`, a list of messages, or `None` to skip the item.
 
+Override the async `ThreadItemConverter.generated_image_to_input()` method to
+customize generated image context, such as resolving private image URLs.
+Return a `Message`, a list of messages, or `None` to skip the item.
+The default converter preserves data URI media types and leaves external URLs unchanged.
+
 ## Installation
 
 ```bash
diff --git a/python/packages/chatkit/agent_framework_chatkit/_converter.py b/python/packages/chatkit/agent_framework_chatkit/_converter.py
index f6501178f9..d878da2f06 100644
--- a/python/packages/chatkit/agent_framework_chatkit/_converter.py
+++ b/python/packages/chatkit/agent_framework_chatkit/_converter.py
@@ -524,6 +524,36 @@ async def structured_input_to_input(self, item: StructuredInputItem) -> Message
         )
         return Message(role="user", contents=[text])
 
+    async def generated_image_to_input(self, item: GeneratedImageItem) -> Message | list[Message] | None:
+        """Convert a generated image to Agent Framework Message(s).
+
+        This method is called internally by `to_agent_input()`. Override this method
+        to customize image conversion, such as resolving a private image URL.
+
+        Args:
+            item: The ChatKit generated image item to convert.
+
+        Returns:
+            A Message with user role, a list of messages, or None to skip.
+
+        Raises:
+            ContentError: If the generated image URL is invalid.
+
+        Note:
+            Use `to_agent_input()` to convert thread items with proper message ordering.
+        """
+        if item.image is None:
+            return None
+
+        uri = item.image.url
+        return Message(
+            role="user",
+            contents=[
+                Content.from_text(text="The following image was generated by the agent."),
+                Content.from_uri(uri=uri, media_type=None if uri.startswith("data:") else "image/*"),
+            ],
+        )
+
     async def _thread_item_to_input_item(
         self,
         item: ThreadItem,
@@ -567,8 +597,8 @@ async def _thread_item_to_input_item(
                 out = self.hidden_context_to_input(item) or []
                 return out if isinstance(out, list) else [out]
             case GeneratedImageItem():
-                # TODO(evmattso): Implement generated image handling in a future PR
-                return []
+                out = await self.generated_image_to_input(item) or []
+                return out if isinstance(out, list) else [out]
             case StructuredInputItem():
                 out = await self.structured_input_to_input(item) or []
                 return out if isinstance(out, list) else [out]
diff --git a/python/packages/chatkit/tests/test_converter.py b/python/packages/chatkit/tests/test_converter.py
index b9889f14db..2633abdda0 100644
--- a/python/packages/chatkit/tests/test_converter.py
+++ b/python/packages/chatkit/tests/test_converter.py
@@ -745,6 +745,118 @@ async def structured_input_to_input(self, item: StructuredInputItem) -> Message
 
         assert exception.value is error
 
+    @pytest.mark.parametrize("url", ["https://example.com/image.png", "https://example.com/asset?signature=sample"])
+    async def test_to_agent_input_with_generated_image(self, converter, url):
+        """Test generated images are preserved as multimodal model context."""
+        from chatkit.types import GeneratedImage, GeneratedImageItem
+
+        input_item = GeneratedImageItem(
+            id="generated_1",
+            thread_id="thread_1",
+            created_at=datetime.now(),
+            type="generated_image",
+            image=GeneratedImage(id="image_1", url=url),
+        )
+
+        result = await converter.to_agent_input(input_item)
+
+        assert len(result) == 1
+        assert result[0].role == "user"
+        assert [content.type for content in result[0].contents] == ["text", "uri"]
+        assert result[0].contents[0].text == "The following image was generated by the agent."
+        assert result[0].contents[1].uri == url
+        assert result[0].contents[1].media_type == "image/*"
+
+    @pytest.mark.parametrize("media_type", ["image/png", "image/jpeg", "image/webp"])
+    async def test_to_agent_input_preserves_generated_image_data_uri_media_type(self, converter, media_type):
+        """Test data URI media types remain available to provider serializers."""
+        from chatkit.types import GeneratedImage, GeneratedImageItem
+
+        uri = f"data:{media_type};base64,YQ=="
+        input_item = GeneratedImageItem(
+            id="generated_1",
+            thread_id="thread_1",
+            created_at=datetime.now(),
+            image=GeneratedImage(id="image_1", url=uri),
+        )
+
+        result = await converter.to_agent_input(input_item)
+
+        assert len(result) == 1
+        assert result[0].role == "user"
+        assert [content.type for content in result[0].contents] == ["text", "data"]
+        assert result[0].contents[1].uri == uri
+        assert result[0].contents[1].media_type == media_type
+
+    async def test_generated_image_to_input_without_image_returns_none(self, converter):
+        """Test unfinished image items are skipped by both conversion entry points."""
+        from chatkit.types import GeneratedImageItem
+
+        input_item = GeneratedImageItem(id="generated_1", thread_id="thread_1", created_at=datetime.now())
+
+        assert await converter.generated_image_to_input(input_item) is None
+        assert await converter.to_agent_input(input_item) == []
+
+    @pytest.mark.parametrize(
+        ("converted", "expected_texts"),
+        [
+            (None, []),
+            ([], []),
+            (Message("user", ["image omitted"]), ["image omitted"]),
+            ([Message("user", ["first"]), Message("user", ["second"])], ["first", "second"]),
+        ],
+        ids=["skip", "empty", "single", "multiple"],
+    )
+    async def test_to_agent_input_uses_generated_image_override(
+        self, converted: Message | list[Message] | None, expected_texts: list[str]
+    ) -> None:
+        """Test async image overrides can skip or expand an item while preserving order."""
+        from chatkit.types import GeneratedImage, GeneratedImageItem, HiddenContextItem
+
+        calls: list[GeneratedImageItem] = []
+
+        class CustomConverter(ThreadItemConverter):
+            async def generated_image_to_input(self, item: GeneratedImageItem) -> Message | list[Message] | None:
+                calls.append(item)
+                return converted
+
+        input_item = GeneratedImageItem(
+            id="generated_1",
+            thread_id="thread_1",
+            created_at=datetime.now(),
+            image=GeneratedImage(id="image_1", url="https://example.com/image.png"),
+        )
+        before = HiddenContextItem(id="before", thread_id="thread_1", created_at=datetime.now(), content="before")
+        after = HiddenContextItem(id="after", thread_id="thread_1", created_at=datetime.now(), content="after")
+
+        result = await CustomConverter().to_agent_input([before, input_item, after])
+
+        assert calls == [input_item]
+        assert [message.text for message in result[1:-1]] == expected_texts
+        assert result[0].text == "<HIDDEN_CONTEXT>before</HIDDEN_CONTEXT>"
+        assert result[-1].text == "<HIDDEN_CONTEXT>after</HIDDEN_CONTEXT>"
+
+    @pytest.mark.parametrize("error", [ValueError("conversion failed"), asyncio.CancelledError()])
+    async def test_to_agent_input_propagates_generated_image_override_errors(self, error: BaseException) -> None:
+        """Test custom image conversion failures and cancellation propagate to the caller."""
+        from chatkit.types import GeneratedImage, GeneratedImageItem
+
+        class FailingConverter(ThreadItemConverter):
+            async def generated_image_to_input(self, item: GeneratedImageItem) -> Message | list[Message] | None:
+                raise error
+
+        input_item = GeneratedImageItem(
+            id="generated_1",
+            thread_id="thread_1",
+            created_at=datetime.now(),
+            image=GeneratedImage(id="image_1", url="https://example.com/image.png"),
+        )
+
+        with pytest.raises(type(error)) as exception:
+            await FailingConverter().to_agent_input(input_item)
+
+        assert exception.value is error
+
     async def test_to_agent_input_dispatches_supported_variants(self, converter):
         """Test thread item dispatch converts supported items and skips unsupported variants."""
         from chatkit.types import (