python/docs/src/user-guide/agentchat-user-guide/tutorial/agents.ipynb
AutoGen AgentChat provides a set of preset Agents, each with variations in how an agent might respond to messages. All agents share the following attributes and methods:
~autogen_agentchat.agents.BaseChatAgent.name: The unique name of the agent.~autogen_agentchat.agents.BaseChatAgent.description: The description of the agent in text.~autogen_agentchat.agents.BaseChatAgent.run: The method that runs the agent given a task as a string or a list of messages, and returns a {py:class}~autogen_agentchat.base.TaskResult. Agents are expected to be stateful and this method is expected to be called with new messages, not complete history.~autogen_agentchat.agents.BaseChatAgent.run_stream: Same as {py:meth}~autogen_agentchat.agents.BaseChatAgent.run but returns an iterator of messages that subclass {py:class}~autogen_agentchat.messages.BaseAgentEvent or {py:class}~autogen_agentchat.messages.BaseChatMessage followed by a {py:class}~autogen_agentchat.base.TaskResult as the last item.See {py:mod}autogen_agentchat.messages for more information on AgentChat message types.
{py:class}~autogen_agentchat.agents.AssistantAgent is a built-in agent that
uses a language model and has the ability to use tools.
{py:class}`~autogen_agentchat.agents.AssistantAgent` is a "kitchen sink" agent
for prototyping and educational purpose -- it is very general.
Make sure you read the documentation and implementation to understand the design choices.
Once you fully understand the design, you may want to implement your own agent.
See [Custom Agent](../custom-agents.ipynb).
from autogen_agentchat.agents import AssistantAgent
from autogen_agentchat.messages import StructuredMessage
from autogen_agentchat.ui import Console
from autogen_ext.models.openai import OpenAIChatCompletionClient
# Define a tool that searches the web for information.
# For simplicity, we will use a mock function here that returns a static string.
async def web_search(query: str) -> str:
"""Find information on the web"""
return "AutoGen is a programming framework for building multi-agent applications."
# Create an agent that uses the OpenAI GPT-4o model.
model_client = OpenAIChatCompletionClient(
model="gpt-4.1-nano",
# api_key="YOUR_API_KEY",
)
agent = AssistantAgent(
name="assistant",
model_client=model_client,
tools=[web_search],
system_message="Use tools to solve tasks.",
)
We can use the {py:meth}~autogen_agentchat.agents.BaseChatAgent.run method to get the agent run on a given task.
# Use asyncio.run(agent.run(...)) when running in a script.
result = await agent.run(task="Find information on AutoGen")
print(result.messages)
The call to the {py:meth}~autogen_agentchat.agents.BaseChatAgent.run method
returns a {py:class}~autogen_agentchat.base.TaskResult
with the list of messages in the {py:attr}~autogen_agentchat.base.TaskResult.messages attribute,
which stores the agent's "thought process" as well as the final response.
It is important to note that {py:meth}`~autogen_agentchat.agents.BaseChatAgent.run`
will update the internal state of the agent -- it will add the messages to the agent's
message history. You can also call {py:meth}`~autogen_agentchat.agents.BaseChatAgent.run`
without a task to get the agent to generate responses given its current state.
Unlike in v0.2 AgentChat, the tools are executed by the same agent directly within
the same call to {py:meth}`~autogen_agentchat.agents.BaseChatAgent.run`.
By default, the agent will return the result of the tool call as the final response.
The {py:class}~autogen_agentchat.agents.AssistantAgent can handle multi-modal input
by providing the input as a {py:class}~autogen_agentchat.messages.MultiModalMessage.
from io import BytesIO
import PIL
import requests
from autogen_agentchat.messages import MultiModalMessage
from autogen_core import Image
# Create a multi-modal message with random image and text.
pil_image = PIL.Image.open(BytesIO(requests.get("https://picsum.photos/300/200").content))
img = Image(pil_image)
multi_modal_message = MultiModalMessage(content=["Can you describe the content of this image?", img], source="user")
img
# Use asyncio.run(...) when running in a script.
result = await agent.run(task=multi_modal_message)
print(result.messages[-1].content) # type: ignore
We can also stream each message as it is generated by the agent by using the
{py:meth}~autogen_agentchat.agents.BaseChatAgent.run_stream method,
and use {py:class}~autogen_agentchat.ui.Console to print the messages
as they appear to the console.
async def assistant_run_stream() -> None:
# Option 1: read each message from the stream (as shown in the previous example).
# async for message in agent.run_stream(task="Find information on AutoGen"):
# print(message)
# Option 2: use Console to print all messages as they appear.
await Console(
agent.run_stream(task="Find information on AutoGen"),
output_stats=True, # Enable stats printing.
)
# Use asyncio.run(assistant_run_stream()) when running in a script.
await assistant_run_stream()
The {py:meth}~autogen_agentchat.agents.BaseChatAgent.run_stream method
returns an asynchronous generator that yields each message generated by the agent,
followed by a {py:class}~autogen_agentchat.base.TaskResult as the last item.
From the messages, you can observe that the assistant agent utilized the web_search tool to
gather information and responded based on the search results.
Large Language Models (LLMs) are typically limited to generating text or code responses. However, many complex tasks benefit from the ability to use external tools that perform specific actions, such as fetching data from APIs or databases.
To address this limitation, modern LLMs can now accept a list of available tool schemas (descriptions of tools and their arguments) and generate a tool call message. This capability is known as Tool Calling or Function Calling and is becoming a popular pattern in building intelligent agent-based applications. Refer to the documentation from OpenAI and Anthropic for more information about tool calling in LLMs.
In AgentChat, the {py:class}~autogen_agentchat.agents.AssistantAgent can use tools to perform specific actions.
The web_search tool is one such tool that allows the assistant agent to search the web for information.
A single custom tool can be a Python function or a subclass of the {py:class}~autogen_core.tools.BaseTool.
On the other hand, a {py:class}~autogen_core.tools.Workbench is a collection of tools that share state and resources.
For how to use model clients directly with tools and workbench, refer to the [Tools](../../core-user-guide/components/tools.ipynb)
and [Workbench](../../core-user-guide/components/workbench.ipynb) sections
in the Core User Guide.
By default, when {py:class}~autogen_agentchat.agents.AssistantAgent executes a tool,
it will return the tool's output as a string in {py:class}~autogen_agentchat.messages.ToolCallSummaryMessage in its response.
If your tool does not return a well-formed string in natural language, you
can add a reflection step to have the model summarize the tool's output,
by setting the reflect_on_tool_use=True parameter in the {py:class}~autogen_agentchat.agents.AssistantAgent constructor.
AutoGen Extension provides a set of built-in tools that can be used with the Assistant Agent.
Head over to the API documentation for all the available tools
under the autogen_ext.tools namespace. For example, you can find the following tools:
~autogen_ext.tools.graphrag: Tools for using GraphRAG index.~autogen_ext.tools.http: Tools for making HTTP requests.~autogen_ext.tools.langchain: Adaptor for using LangChain tools.~autogen_ext.tools.mcp: Tools and workbench for using Model Chat Protocol (MCP) servers.The {py:class}~autogen_agentchat.agents.AssistantAgent automatically
converts a Python function into a {py:class}~autogen_core.tools.FunctionTool
which can be used as a tool by the agent and automatically generates the tool schema
from the function signature and docstring.
The web_search_func tool is an example of a function tool.
The schema is automatically generated.
from autogen_core.tools import FunctionTool
# Define a tool using a Python function.
async def web_search_func(query: str) -> str:
"""Find information on the web"""
return "AutoGen is a programming framework for building multi-agent applications."
# This step is automatically performed inside the AssistantAgent if the tool is a Python function.
web_search_function_tool = FunctionTool(web_search_func, description="Find information on the web")
# The schema is provided to the model during AssistantAgent's on_messages call.
web_search_function_tool.schema
The {py:class}~autogen_agentchat.agents.AssistantAgent can also use tools that are
served from a Model Context Protocol (MCP) server
using {py:func}~autogen_ext.tools.mcp.McpWorkbench.
from autogen_agentchat.agents import AssistantAgent
from autogen_agentchat.messages import TextMessage
from autogen_ext.models.openai import OpenAIChatCompletionClient
from autogen_ext.tools.mcp import McpWorkbench, StdioServerParams
# Get the fetch tool from mcp-server-fetch.
fetch_mcp_server = StdioServerParams(command="uvx", args=["mcp-server-fetch"])
# Create an MCP workbench which provides a session to the mcp server.
async with McpWorkbench(fetch_mcp_server) as workbench: # type: ignore
# Create an agent that can use the fetch tool.
model_client = OpenAIChatCompletionClient(model="gpt-4.1-nano")
fetch_agent = AssistantAgent(
name="fetcher", model_client=model_client, workbench=workbench, reflect_on_tool_use=True
)
# Let the agent fetch the content of a URL and summarize it.
result = await fetch_agent.run(task="Summarize the content of https://en.wikipedia.org/wiki/Seattle")
assert isinstance(result.messages[-1], TextMessage)
print(result.messages[-1].content)
# Close the connection to the model client.
await model_client.close()
Any {py:class}~autogen_agentchat.agents.BaseChatAgent can be used as a tool
by wrapping it in a {py:class}~autogen_agentchat.tools.AgentTool.
This allows for a dynamic, model-driven multi-agent workflow where
the agent can call other agents as tools to solve tasks.
Some models support parallel tool calls, which can be useful for tasks that require multiple tools to be called simultaneously.
By default, if the model client produces multiple tool calls, {py:class}~autogen_agentchat.agents.AssistantAgent
will call the tools in parallel.
You may want to disable parallel tool calls when the tools have side effects that may interfere with each other, or, when agent behavior needs to be consistent across different models. This should be done at the model client level.
When using {py:class}`~autogen_agentchat.tools.AgentTool` or {py:class}`~autogen_agentchat.tools.TeamTool`,
you **must** disable parallel tool calls to avoid concurrency issues.
These tools cannot run concurrently as agents and teams maintain internal state
that would conflict with parallel execution.
For {py:class}~autogen_ext.models.openai.OpenAIChatCompletionClient and {py:class}~autogen_ext.models.openai.AzureOpenAIChatCompletionClient,
set parallel_tool_calls=False to disable parallel tool calls.
model_client_no_parallel_tool_call = OpenAIChatCompletionClient(
model="gpt-4o",
parallel_tool_calls=False, # type: ignore
)
agent_no_parallel_tool_call = AssistantAgent(
name="assistant",
model_client=model_client_no_parallel_tool_call,
tools=[web_search],
system_message="Use tools to solve tasks.",
)
One model call followed by one tool call or parallel tool calls
is a single tool iteration.
By default, the {py:class}~autogen_agentchat.agents.AssistantAgent will
execute at most one iteration.
The agent can be configured to execute multiple iterations until the model
stops generating tool calls or the maximum number of iterations is reached.
You can control the maximum number of iterations by setting the max_tool_iterations parameter
in the {py:class}~autogen_agentchat.agents.AssistantAgent constructor.
agent_loop = AssistantAgent(
name="assistant_loop",
model_client=model_client_no_parallel_tool_call,
tools=[web_search],
system_message="Use tools to solve tasks.",
max_tool_iterations=10, # At most 10 iterations of tool calls before stopping the loop.
)
Structured output allows models to return structured JSON text with pre-defined schema provided by the application. Different from JSON-mode, the schema can be provided as a Pydantic BaseModel class, which can also be used to validate the output.
Once you specify the base model class in the output_content_type parameter
of the {py:class}~autogen_agentchat.agents.AssistantAgent constructor,
the agent will respond with a {py:class}~autogen_agentchat.messages.StructuredMessage
whose content's type is the type of the base model class.
This way, you can integrate agent's response directly into your application and use the model's output as a structured object.
When the `output_content_type` is set, it by default requires the agent to reflect on the tool use
and return the a structured output message based on the tool call result.
You can disable this behavior by setting `reflect_on_tool_use=False` explictly.
Structured output is also useful for incorporating Chain-of-Thought reasoning in the agent's responses. See the example below for how to use structured output with the assistant agent.
from typing import Literal
from pydantic import BaseModel
# The response format for the agent as a Pydantic base model.
class AgentResponse(BaseModel):
thoughts: str
response: Literal["happy", "sad", "neutral"]
# Create an agent that uses the OpenAI GPT-4o model.
model_client = OpenAIChatCompletionClient(model="gpt-4o")
agent = AssistantAgent(
"assistant",
model_client=model_client,
system_message="Categorize the input as happy, sad, or neutral following the JSON format.",
# Define the output content type of the agent.
output_content_type=AgentResponse,
)
result = await Console(agent.run_stream(task="I am happy."))
# Check the last message in the result, validate its type, and print the thoughts and response.
assert isinstance(result.messages[-1], StructuredMessage)
assert isinstance(result.messages[-1].content, AgentResponse)
print("Thought: ", result.messages[-1].content.thoughts)
print("Response: ", result.messages[-1].content.response)
await model_client.close()
You can stream the tokens generated by the model client by setting model_client_stream=True.
This will cause the agent to yield {py:class}~autogen_agentchat.messages.ModelClientStreamingChunkEvent messages
in {py:meth}~autogen_agentchat.agents.BaseChatAgent.run_stream.
The underlying model API must support streaming tokens for this to work. Please check with your model provider to see if this is supported.
model_client = OpenAIChatCompletionClient(model="gpt-4o")
streaming_assistant = AssistantAgent(
name="assistant",
model_client=model_client,
system_message="You are a helpful assistant.",
model_client_stream=True, # Enable streaming tokens.
)
# Use an async function and asyncio.run() in a script.
async for message in streaming_assistant.run_stream(task="Name two cities in South America"): # type: ignore
print(message)
You can see the streaming chunks in the output above. The chunks are generated by the model client and are yielded by the agent as they are received. The final response, the concatenation of all the chunks, is yielded right after the last chunk.
{py:class}~autogen_agentchat.agents.AssistantAgent has a model_context
parameter that can be used to pass in a {py:class}~autogen_core.model_context.ChatCompletionContext
object. This allows the agent to use different model contexts, such as
{py:class}~autogen_core.model_context.BufferedChatCompletionContext to
limit the context sent to the model.
By default, {py:class}~autogen_agentchat.agents.AssistantAgent uses
the {py:class}~autogen_core.model_context.UnboundedChatCompletionContext
which sends the full conversation history to the model. To limit the context
to the last n messages, you can use the {py:class}~autogen_core.model_context.BufferedChatCompletionContext.
To limit the context by token count, you can use the
{py:class}~autogen_core.model_context.TokenLimitedChatCompletionContext.
from autogen_core.model_context import BufferedChatCompletionContext
# Create an agent that uses only the last 5 messages in the context to generate responses.
agent = AssistantAgent(
name="assistant",
model_client=model_client,
tools=[web_search],
system_message="Use tools to solve tasks.",
model_context=BufferedChatCompletionContext(buffer_size=5), # Only use the last 5 messages in the context.
)
The following preset agents are available:
~autogen_agentchat.agents.UserProxyAgent: An agent that takes user input returns it as responses.~autogen_agentchat.agents.CodeExecutorAgent: An agent that can execute code.~autogen_ext.agents.openai.OpenAIAssistantAgent: An agent that is backed by an OpenAI Assistant, with ability to use custom tools.~autogen_ext.agents.web_surfer.MultimodalWebSurfer: A multi-modal agent that can search the web and visit web pages for information.~autogen_ext.agents.file_surfer.FileSurfer: An agent that can search and browse local files for information.~autogen_ext.agents.video_surfer.VideoSurfer: An agent that can watch videos for information.Having explored the usage of the {py:class}~autogen_agentchat.agents.AssistantAgent, we can now proceed to the next section to learn about the teams feature in AgentChat.