# HITL tool is resuming perfectly in local dev but not it langgraph cloud

**URL:** <https://forum.langchain.com/t/hitl-tool-is-resuming-perfectly-in-local-dev-but-not-it-langgraph-cloud/2278>\
**Category:** LangGraph\
**Tags:** python-help\
**Created:** [November 21, 2025, 6:17am UTC](https://forum.langchain.com/t/hitl-tool-is-resuming-perfectly-in-local-dev-but-not-it-langgraph-cloud/2278 "2025-11-21T06:17:14Z")\
**Posts on this page:** 1\
**Showing post:** 1

<div class="post-metadata">

**Author:** ![pranavtreppan](https://yyz1.discourse-cdn.com/flex007/user_avatar/forum.langchain.com/pranavtreppan/32/1694_2.png) [@pranavtreppan](https://forum.langchain.com/u/pranavtreppan)\
**Post date:** [November 21, 2025, 6:17am UTC](https://forum.langchain.com/t/hitl-tool-is-resuming-perfectly-in-local-dev-but-not-it-langgraph-cloud/2278/1 "2025-11-21T06:17:14Z")

</div>

The project works fine when LangGraph is built and run locally, but the issues appear only when deploying in the cloud environment. When I point in LANGGRAPH\_URL=‘[http://127.0.0.1:2024](http://127.0.0.1:2024)’, its working fine but in langgraph cloud its not working when resuming.

Here is the code snippet:

`from typing import Callable, Optional

from langchain\_core.tools import BaseTool, tool as create\_tool  
from langchain\_core.runnables import RunnableConfig  
from langgraph.types import interrupt, Command  
from langgraph.prebuilt.interrupt import HumanInterruptConfig, HumanInterrupt  
from langgraph\_app.states.file\_data import FileData

def add\_human\_in\_the\_loop(  
tool: Callable | BaseTool,  
\*,  
interrupt\_config: HumanInterruptConfig = None,  
) → BaseTool:  
“”“Wrap a tool to support human-in-the-loop review.”“”  
if not isinstance(tool, BaseTool):  
tool = create\_tool(tool)

```
if interrupt_config is None:
    interrupt_config = {
        "allow_accept": True,
        "allow_edit": True,
        "allow_respond": True,
    }

@create_tool(
    tool.name,
    description=tool.description,
    args_schema=tool.args_schema
)
async def call_tool_with_interrupt(config: RunnableConfig, question: str, field_asked_for: Optional[str]=None, **tool_input):
    tool_call_id = (
            tool_input.pop("tool_call_id", None)
            or tool_input.pop("_tool_call_id", None)
    )
    request: HumanInterrupt = {
        "action_request": {
            "action": tool.name,
            "args": {"question": question, "field_asked_for": field_asked_for}
        },
        "config": interrupt_config,
        "description": "Please review the tool call"
    }
    response = interrupt([request])
    text = ""
    imgs = []
    files = []
    for humanmsg in response:
        obj_list=humanmsg.content
        for obj in obj_list:
            # if obj.get("type")=="text":
            # text += obj.get("content", "")
            if obj.get("type") == "text":
                text_value = obj.get("content") or obj.get("text") or ""
                text += text_value
            elif obj.get("type") == "image_url":
                imgs.append(obj.get("image_url").get("url"))
            elif obj.get("type") == "file":
                file_content = obj.get("content", {})
                files.append(FileData(file_name=file_content.get("filename", ""), file_data=file_content.get("file_data", "")))

    tool_input["user_text_response"] = text
    tool_input["user_img_response"] = imgs
    tool_input["user_file_response"] = files
    print(f"----------->>{tool_call_id}")
    # tool_input["tool_call_id"] = tool_call_id

    tool_input["question"] = question
    tool_input["field_asked_for"] = field_asked_for
    tool_response = await tool.ainvoke(tool_input, config)

    return tool_response

return call_tool_with_interrupt`

```

`  
@tool  
async def need\_human\_input(question: str, user\_text\_response: Optional[str], state: Annotated[State, InjectedState],  
tool\_call\_id: Annotated[str, InjectedToolCallId]):  
“”"  
Use this tool when you need to ask user about product or clarification.  
Example: “Which product are you referring to?” or “Is it shower cubicle or sliding door?”  
“”"

```
new_msgs = [
    ToolMessage(
        f"Waiting for response", tool_call_id=tool_call_id,
        name='ask_query_to_human'),
    AIMessage(content=question),

]
user_response = []
if user_text_response:
    user_response.append({"type": "text", "text": user_text_response})

state.messages = new_msgs + [HumanMessage(content=user_response)]
return Command(update=state)

```

` `  
def fetch\_supervisor\_tools():  
TOOLS: List[Callable[…, Any]] = [  
add\_human\_in\_the\_loop(need\_human\_input),  
]  
return TOOLS`

ERROR: When tool includes an InjectedToolCallId argument, tool must always be invoked with a full model ToolCall of the form: {‘args’: {…}, ‘name’: ‘…’, ‘type’: ‘tool\_call’, ‘tool\_call\_id’: ‘…’}

This above error is not raised in local langgraph dev but raised in langgraph cloud issue. Is it due to InjectedToolCallID param in the tool?

---

_[View the full topic](https://forum.langchain.com/t/hitl-tool-is-resuming-perfectly-in-local-dev-but-not-it-langgraph-cloud/2278)._