> ## Documentation Index
> Fetch the complete documentation index at: https://docs.orbitrage.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# LangGraph

> Use Orbitrage with LangGraph agents.

LangGraph uses LangChain chat models under the hood — point them at Orbitrage
and your whole graph is routed and traced. Each node's call shows up in the
workflow flow graph.

```python theme={null}
import os, orbitrage
orbitrage.init(os.environ["ORBITRAGE_API_KEY"], user_id="customer_42")

from langgraph.prebuilt import create_react_agent
from langchain_openai import ChatOpenAI

llm = ChatOpenAI(
    base_url="https://api.orbitrage.ai/v1",
    api_key=os.environ["ORBITRAGE_API_KEY"],
    model="minimax-m3",
)

agent = create_react_agent(llm, tools=[])     # add your tools
agent.invoke({"messages": [{"role": "user", "content": "What's 2+2?"}]})
```

<Tip>
  All calls in one process share a run id, so a multi-step graph appears as a
  single run you can replay node by node in the dashboard.
</Tip>