Implementation:Microsoft Agent framework Agent Init

Property	Value
Implementation Name	Agent Init
SDK	Microsoft Agent Framework
Repository	agent-framework
Source File	`python/packages/core/agent_framework/_agents.py`
Line Range	L1377-1426
Import	`from agent_framework import Agent`
Type	Class constructor (`__init__`)

Overview

The Agent.__init__ method is the constructor for the central Agent class in the Microsoft Agent Framework. It binds together a chat client, system instructions, and optional tools into a configured agent instance ready for execution via run(). The Agent class extends BaseAgent[OptionsCoT] and is generic over OptionsCoT, enabling type-safe propagation of model-specific options throughout the agent lifecycle.

Code Reference

Source Location

Property	Value
File	`python/packages/core/agent_framework/_agents.py`
Class	`Agent`
Base Classes	`BaseAgent[OptionsCoT]`, `Generic[OptionsCoT]`
Lines	1377-1426

Signature

class Agent(BaseAgent[OptionsCoT], Generic[OptionsCoT]):
    def __init__(
        self,
        client: SupportsChatGetResponse[OptionsCoT],
        instructions: str = "",
        *,
        id: str | None = None,
        name: str | None = None,
        description: str | None = None,
        tools: FunctionTool | Callable | Sequence[FunctionTool | Callable] | None = None,
        default_options: OptionsCoT | None = None,
        chat_message_store_factory: ChatMessageStoreFactory | None = None,
        context_provider: ContextProvider | None = None,
        middleware: Sequence[
            AgentMiddleware
            | AgentMiddlewareCallable
            | ChatMiddleware
            | ChatMiddlewareCallable
            | FunctionMiddleware
            | FunctionMiddlewareCallable
        ] | None = None,
        function_invocation_configuration: FunctionInvocationConfiguration | None = None,
        **kwargs: Any,
    ) -> None:

Import Statement

from agent_framework import Agent

I/O Contract

Inputs

Parameter	Type	Default	Description
`client`	`SupportsChatGetResponse[OptionsCoT]`	(required)	The chat client that provides LLM inference. Must implement the `SupportsChatGetResponse` protocol.
`instructions`	`str`	`""`	System prompt string that defines the agent's behavior, persona, and constraints.
`id`	None	`None`	Optional unique identifier for the agent instance.
`name`	None	`None`	Optional human-readable name for identification and tracing.
`description`	None	`None`	Optional description of the agent's purpose and capabilities.
`tools`	Callable \| Sequence[FunctionTool \| Callable] \| None	`None`	Tools the agent can invoke. Accepts a single tool, a plain callable, or a sequence of either. `Callable` values are automatically wrapped into `FunctionTool` instances.
`default_options`	None	`None`	Model-specific options (e.g., temperature, max tokens) generic over the options type parameter.
`chat_message_store_factory`	None	`None`	Factory for creating message stores that persist or cache conversation history.
`context_provider`	None	`None`	Injectable component that supplies additional runtime context during agent execution.
`middleware`	AgentMiddlewareCallable \| ChatMiddleware \| ChatMiddlewareCallable \| FunctionMiddleware \| FunctionMiddlewareCallable] \| None	`None`	Middleware pipeline that intercepts and transforms agent, chat, or function invocation behavior.
`function_invocation_configuration`	None	`None`	Configuration controlling how tool functions are invoked (e.g., parallel execution, error handling).
`**kwargs`	`Any`		Additional keyword arguments forwarded to the base class.

Output

Type	Description
`Agent[OptionsCoT]`	A fully configured agent instance ready for execution via `agent.run()` or within an async context manager.

Usage Examples

Minimal Agent

from agent_framework import Agent
from agent_framework.openai import OpenAIResponsesClient

client = OpenAIResponsesClient()
agent = Agent(
    client=client,
    instructions="You are a helpful assistant.",
    name="assistant",
)

Agent with Tools

from agent_framework import Agent
from agent_framework.openai import OpenAIResponsesClient

client = OpenAIResponsesClient()
agent = Agent(
    client=client,
    instructions="You are a helpful assistant.",
    name="assistant",
    tools=[get_weather],
)

Agent with Middleware

from agent_framework import Agent
from agent_framework.openai import OpenAIResponsesClient

client = OpenAIResponsesClient()
agent = Agent(
    client=client,
    instructions="You are a helpful assistant.",
    name="assistant",
    tools=[get_weather],
    middleware=[logging_middleware, auth_middleware],
)

Factory Method Alternative

agent = client.as_agent(
    instructions="You are a helpful assistant.",
    tools=[get_weather],
)

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment