Observing agents built with LangGraph

Make sure that the following prerequisites are met.

• Python 3.8 or later is installed.

• An LLM API key such as Anthropic, OpenAI, or any LLM provider supported by LangGraph.

• A search API key (optional, for search tool functionality - Tavily or DuckDuckGo).

• Instana is configured for your application, see Getting started.

To instrument your LangGraph application with Instana, complete the following steps:

1. Install the required packages for LangGraph and OpenLLMetry.

   pip install langgraph langchain-anthropic duckduckgo-search traceloop-sdk

2. Set your LLM provider and search tool credentials.

   export ANTHROPIC_API_KEY="<your-api-key>"
   export TAVILY_API_KEY="<your-tavily-api-key>"

   Note:

   Replace with your LLM provider's environment variable if using a different provider (e.g., OPENAI_API_KEY for OpenAI). The Tavily API key is optional if you're using a different search tool instead.

3. Create your LangGraph application.

   A multi-agent workflow with a researcher and a writer is shown in the following example:

   import os
   from typing import Literal
   from langchain_anthropic import ChatAnthropic
   from langchain_community.tools import DuckDuckGoSearchRun
   from langchain_core.messages import BaseMessage, HumanMessage
   from langgraph.prebuilt import create_react_agent
   from langgraph.graph import MessagesState, StateGraph, START, END
   from langgraph.types import Command
   from traceloop.sdk import Traceloop
   from traceloop.sdk.decorators import workflow
   
   # Initialize OpenLLMetry
   Traceloop.init(app_name="langgraph_demo")
   
   # Define search tool
   search_tool = DuckDuckGoSearchRun()
   
   # Define LLM
   llm = ChatAnthropic(
       model="claude-3-5-sonnet-20241022",
       temperature=0.7,
       max_tokens=600
   )
   
   def get_next_node(last_message: BaseMessage, goto: str):
       if not last_message.content:
           return END
       return goto
   
   # Research agent and node
   research_agent = create_react_agent(
       llm,
       tools=[search_tool],
       state_modifier="You are a research assistant. Your job is to conduct thorough research."
   )
   
   def research_node(state: MessagesState) -> Command[Literal["writer", END]]:
       result = research_agent.invoke(state)
       goto = get_next_node(result["messages"][-1], "writer")
       result["messages"][-1] = HumanMessage(
           content=result["messages"][-1].content,
           name="researcher"
       )
       return Command(
           update={"messages": result["messages"]},
           goto=goto
       )
   
   # Writer agent and node
   writer_agent = create_react_agent(
       llm,
       tools=[search_tool],
       state_modifier="You are a blog writer. Your task is to create a well-structured blog post."
   )
   
   def writer_node(state: MessagesState) -> Command[Literal[END]]:
       result = writer_agent.invoke(state)
       result["messages"][-1] = HumanMessage(
           content=result["messages"][-1].content,
           name="writer"
       )
       return Command(
           update={"messages": result["messages"]},
           goto=END
       )
   
   # Define workflow graph
   workflow = StateGraph(MessagesState)
   workflow.add_node("researcher", research_node)
   workflow.add_node("writer", writer_node)
   workflow.add_edge(START, "researcher")
   graph = workflow.compile()
   
   # Execute workflow
   @workflow(name="langgraph_multi_agent_workflow")
   def run_workflow():
       events = graph.stream(
           {
               "messages": [
                   (
                       "user",
                       "First, research AI trends in 2025. "
                       "Then, based on that research, create a well-structured blog post."
                   )
               ]
           },
           {"recursion_limit": 150}
       )
   
       # Print workflow execution steps
       for event in events:
           print(event)
           print("----")
   
   if __name__ == "__main__":
       run_workflow()

4. Run your LangGraph application.

   python langgraph_app.py

   The application executes the multi-agent workflow and displays the execution steps:

   

5. View data in Instana.

   After running your application, data will appear in the Instana Gen AI observability dashboard:

   

   The dashboard shows:

   • Multi-agent workflow execution

   • State transitions between agents

   • Tool invocations (search operations)

   • LLM interactions with latency and token usage

   • Conditional routing decisions

For common issues such as traces not appearing or connection errors, see Troubleshooting.

• Explore AI agent frameworks supported by Instana

• Learn about cost calculation for your LLM usage

• Set up alerts for your LangGraph workflows