Quickstart¶

This guide walks you through deploying a LangGraph agent as Azure Functions HTTP endpoints in under 5 minutes.

Prerequisites¶

Python 3.10+
Azure Functions Core Tools installed
azure-functions-langgraph installed (see Installation)

Step 1: Define your graph state¶

from typing_extensions import TypedDict


class AgentState(TypedDict):
    messages: list[dict[str, str]]

Step 2: Define node functions¶

def chat(state: AgentState) -> dict:
    user_msg = state["messages"][-1]["content"]
    return {
        "messages": state["messages"] + [
            {"role": "assistant", "content": f"Echo: {user_msg}"}
        ]
    }

Step 3: Build and compile the graph¶

from langgraph.graph import END, START, StateGraph

builder = StateGraph(AgentState)
builder.add_node("chat", chat)
builder.add_edge(START, "chat")
builder.add_edge("chat", END)

graph = builder.compile()

Step 4: Register with LangGraphApp¶

from azure_functions_langgraph import LangGraphApp

app = LangGraphApp()
app.register(graph=graph, name="echo_agent")

Step 5: Export the function app¶

In your function_app.py:

from azure_functions_langgraph import LangGraphApp
from langgraph.graph import END, START, StateGraph
from typing_extensions import TypedDict


class AgentState(TypedDict):
    messages: list[dict[str, str]]


def chat(state: AgentState) -> dict:
    user_msg = state["messages"][-1]["content"]
    return {
        "messages": state["messages"] + [
            {"role": "assistant", "content": f"Echo: {user_msg}"}
        ]
    }


builder = StateGraph(AgentState)
builder.add_node("chat", chat)
builder.add_edge(START, "chat")
builder.add_edge("chat", END)
graph = builder.compile()

langgraph_app = LangGraphApp()
langgraph_app.register(graph=graph, name="echo_agent")

app = langgraph_app.function_app

Step 6: Run locally¶

func start

Step 7: Test the endpoints¶

Invoke the agent:

curl -X POST http://localhost:7071/api/graphs/echo_agent/invoke \
  -H "Content-Type: application/json" \
  -d '{"input": {"messages": [{"role": "human", "content": "Hello!"}]}}'

Response:

{
    "output": {
        "messages": [
            {"role": "human", "content": "Hello!"},
            {"role": "assistant", "content": "Echo: Hello!"}
        ]
    }
}

Stream the agent:

curl -X POST http://localhost:7071/api/graphs/echo_agent/stream \
  -H "Content-Type: application/json" \
  -d '{"input": {"messages": [{"role": "human", "content": "Hello!"}]}, "stream_mode": "values"}'

Check health:

curl http://localhost:7071/api/health

Response:

{
    "status": "ok",
    "graphs": [
        {"name": "echo_agent", "description": null, "has_checkpointer": false}
    ]
}

Next steps¶

Configuration — configure auth levels and graph options
Usage Guide — detailed endpoint reference with all request/response formats
Examples — more complete examples