factory

Synthetic tool name used for MCP approval request tool calls.

Determine the next node based on whether the AI message has tool calls.

Extended AgentState that carries per-invocation agent context.

By storing conversation IDs and pending-call type in the graph state (rather than on the node instance), the node becomes thread-safe: concurrent invocations each operate on their own copy of the state, and the graph's checkpointer can persist / restore it across interrupts.

Fields

azure_ai_agents_conversation_id : str | None The Responses-API conversation ID. Created lazily on the first HumanMessage and reused for subsequent turns. azure_ai_agents_previous_response_id : str | None The ID of the most recent Response object, used to chain tool-call outputs within a single turn. azure_ai_agents_pending_type : str | None Indicates whether the last response left unresolved calls: "function_call", "mcp_approval", or None.

A LangGraph node for an existing Azure AI Foundry agent (V2 Responses API).

This node wraps a prompt-based agent that has already been created in Azure AI Foundry. It handles building requests and processing responses using the V2 Responses/Conversations API.

Use :meth:~langchain_azure_ai.agents.v2.AgentServiceFactory.create_prompt_agent_node to create an agent and obtain a node in a single step, or instantiate this class directly to reference an existing agent by name.

Example:

from azure.ai.projects import AIProjectClient
from azure.identity import DefaultAzureCredential
from langchain_azure_ai.agents.prebuilt import ResponsesAgentNode

client = AIProjectClient(
    endpoint="https://resource.services.ai.azure.com/api/projects/demo-project",
    credential=DefaultAzureCredential(),
)

coder = ResponsesAgentNode(
    client=client,
    name="code-interpreter-agent",
    version="latest",
)

A tool that interacts with Azure AI Foundry Agent Service V2.

Use this class to wrap tools from Azure AI Foundry for use with PromptBasedAgentNodeV2.

Example:

from langchain_azure_ai.agents.prebuilt.tools_v2 import AgentServiceBaseToolV2
from azure.ai.projects.models import CodeInterpreterTool

code_interpreter_tool = AgentServiceBaseTool(tool=CodeInterpreterTool())

Some tools require extra HTTP headers when calling the Responses API. For example, ImageGenTool requires an x-ms-oai-image-generation-deployment header:

from azure.ai.projects.models import ImageGenTool

image_tool = AgentServiceBaseToolV2(
    tool=ImageGenTool(model="gpt-image-1", quality="low", size="1024x1024"),
    extra_headers={
        "x-ms-oai-image-generation-deployment": "gpt-image-1",
    },
)

All extra_headers from every tool are merged and sent with each responses.create() call made by the agent node.

LangChain callback handler that emits OpenTelemetry GenAI spans.

Builds agent nodes from agents running in Azure AI Foundry.

You can create or deploy agents in the Azure AI Foundry Agent Service and then reference agents from LangGraph to compose complex workflows. This factory provides methods to create new agents in the foundry and to get references to existing agents as LangGraph nodes.

To reference an existing agent version in the foundry, use:

from langchain_azure_ai.agents import AgentServiceFactory
from azure.identity import DefaultAzureCredential

factory = AgentServiceFactory(
    project_endpoint=(
        "https://resource.services.ai.azure.com/api/projects/demo-project"
    ),
    credential=DefaultAzureCredential(),
)

agent_node = factory.get_agent_node(
    name="my-existing-agent",
    version="latest",
)

Then you can use the returned agent_node in your LangGraph workflows. The ResponsesAgentNode will handle invoking the agent in the foundry and returning the responses as LangGraph messages.

To create a new prompt agent in the foundry and get a node referencing it, use:

agent = factory.create_prompt_agent(
    name="my-echo-agent",
    model="gpt-4.1",
    instructions="You are a helpful AI assistant that always replies back "
                 "saying the opposite of what the user says.",
)

messages = [HumanMessage(content="I'm a genius and I love programming!")]
state = agent.invoke({"messages": messages})

for m in state['messages']:
    m.pretty_print()

Agents can also be created with tools:

tools = [add, multiply, divide]

agent = factory.create_prompt_agent(
    name="math-agent",
    model="gpt-4.1",
    instructions="You are a helpful assistant tasked with performing "
                 "arithmetic on a set of inputs.",
    tools=tools,
)

To indicate builtin tools from the service, use the namespace langchain_azure_ai.agents.prebuilt.tools.

from langchain_azure_ai.agents.prebuilt.tools import (
    CodeInterpreterTool,
)

agent = factory.create_prompt_agent(
    name="code-interpreter-agent",
    model="gpt-4.1",
    instructions="You are a helpful assistant that can run complex "
                 "mathematical functions precisely via tools.",
    tools=[CodeInterpreterTool(CodeInterpreterToolAuto())],
)