# Information is not passing in Supervisor agent **URL:** \ **Category:** LangGraph\ **Tags:** python-help\ **Created:** [August 7, 2025, 12:15pm UTC](https://forum.langchain.com/t/information-is-not-passing-in-supervisor-agent/997 "2025-08-07T12:15:22Z")\ **Posts on this page:** 3\ **Page:** 1
**Author:** ![IamExperimenting](https://yyz1.discourse-cdn.com/flex007/user_avatar/forum.langchain.com/iamexperimenting/32/279_2.png) [@IamExperimenting](https://forum.langchain.com/u/IamExperimenting)\ **Post date:** [August 7, 2025, 12:15pm UTC](https://forum.langchain.com/t/information-is-not-passing-in-supervisor-agent/997/1 "2025-08-07T12:15:22Z")
I’m building data analyst agent using Supervisor architecture. I want to provide multi-step task to supervisor, supervisor wants to call relevant agents and tools to complete the task. ```auto from typing import Annotated import os, pandas as pd, json, numpy as np from dotenv import load_dotenv load_dotenv() import matplotlib.pyplot as plt from langchain_openai import AzureChatOpenAI from langgraph.graph import StateGraph, START, END, MessagesState from langgraph.graph.message import add_messages from langgraph.checkpoint.memory import MemorySaver from langchain_core.tools import InjectedToolCallId, tool from langchain_core.messages import ToolMessage from langgraph.prebuilt import create_react_agent from langchain.core.messages import convert_to_messages memory = MemorySaver() class CustomState(MessagesState): input_data : pd.DataFrame preprocessed_data : pd.DataFrame def read_data_func(csv_path: str, tool_call_id: Annotated[str, InjectedToolCallId]): """This function read a csv file from a given location.""" df = pd.read_csv(csv_path) return Command(update={"messages:[ToolMessage(df.to_json(), tool_call_id=tool_call_id)], "input_data" : df.to_json() }) def preprocess_func(state: MessagesState, tool_call_id: Annotated[str, InjectedToolCallId]): """This function preprocess a given dataframe.""" state_dict = json.loads(state['input_data']) df = pd.DataFrames(state_dict) df = df[['col1', 'col2']] return Command(update={"messages:[ToolMessage(df.to_json(), tool_call_id=tool_call_id)], "input_data" : df.to_json() }) preprocess_agent = create_react_agent( model="openai:gpt-4.1", tools=[read_data_func, preprocess_func], prompt=( "You are a preprocessor agent.\n\n" "INSTRUCTIONS:\n" "- Assist ONLY with preprocessing-related tasks\n" "- After you're done with your tasks, respond to the supervisor directly\n" "- Respond ONLY with the results of your work, do NOT include ANY other text." ), name="preprocess_agent", ) def graph_gen_func(state: MessagesState, tool_call_id: Annotated[str, InjectedToolCallId]): state_dict = json.loads(state['input_data']) df = pd.DataFrames(state_dict) df.set_index('time_stamp', inplace=True) plt.figure(figsize=(12,6)) plt.plot(df.index, df['data_points'],marker='o') plt.xlabel('timestamp') plt.ylabel('data_points') plt.title('timeseries graph') plt.grid(True) plt.tight_layout() plt.show() graph_generation_agent = create_react_agent( model="openai:gpt-4.1", tools=[graph_gen_func], prompt=( "You are a graph generation agent.\n\n" "INSTRUCTIONS:\n" "- Assist ONLY with garph plotting-related tasks\n" "- After you're done with your tasks, respond to the supervisor directly\n" "- Respond ONLY with the results of your work, do NOT include ANY other text." ), name="graph_generation_agent ", ) from langgraph_supervisor import create_supervisor supervisor = create_superviso( model="openai:gpt-4.1", agents=[preprocess_agent, graph_generation_agent], prompt=( "You are a supervisor managing two agent.\n\n" "- a preprocess agent. Assign preprocessing-related tasks to this agent\n" "- a graph generation agent. Assign graph plotting-related tasks to this agent\n" "Assign work to one agent at a time, do not call agents in parallel.\n" "Do not do any work yourself." ), add_handoff_back_messages=True, output_mode = "full_history" ).compile() for chunk in supervisor.stream( { "messages": [ { "role": "user", "content": "can you please read the file ../data/income_data.csv?", } ] }, ): pretty_print_messages(chunk, last_message=True) final_message_history = chunk["supervisor"]["messages"] for chunk in supervisor.stream( { "messages": [ { "role": "user", "content": "can you please preprocess the given data?", } ] }, ): pretty_print_messages(chunk, last_message=True) final_message_history = chunk["supervisor"]["messages"] for chunk in supervisor.stream( { "messages": [ { "role": "user", "content": "can you please plot graph using preprocessed data?", } ] }, ): pretty_print_messages(chunk, last_message=True) final_message_history = chunk["supervisor"]["messages"] ``` Here, I’m not able to pass information from one tool to other tools and from one agent to other agent. can you please help me? @marco ---
**Author:** ![IamExperimenting](https://yyz1.discourse-cdn.com/flex007/user_avatar/forum.langchain.com/iamexperimenting/32/279_2.png) [@IamExperimenting](https://forum.langchain.com/u/IamExperimenting)\ **Post date:** [August 12, 2025, 1:21pm UTC](https://forum.langchain.com/t/information-is-not-passing-in-supervisor-agent/997/2 "2025-08-12T13:21:50Z")
hi @marco can you please help me to fix this? ---
**Author:** ![MartijnLeplae](https://yyz1.discourse-cdn.com/flex007/user_avatar/forum.langchain.com/martijnleplae/32/664_2.png) [@MartijnLeplae](https://forum.langchain.com/u/MartijnLeplae)\ **Post date:** [August 20, 2025, 2:52pm UTC](https://forum.langchain.com/t/information-is-not-passing-in-supervisor-agent/997/3 "2025-08-20T14:52:12Z")
Could you share traces of LangSmith when executing this code? I think a good check would be to see if the output of the first tool-call is still in the message history when the second message streamed. At first sight I would think you still need to provide the `final_message_history` to the `invoke` and not only the new message.