agent-framework

What it is and what it does

Agent Framework is Microsoft's Python library for building and coordinating AI agents that interact with large language models. It provides a high-level Agent class that wraps OpenAI and Azure OpenAI chat clients, allowing you to define agent behavior through instructions, attach tools for function calling, and compose multiple agents into orchestrated workflows. The framework handles async execution, message passing, and agent-to-agent handoff patterns out of the box.

The package is designed for both simple single-agent tasks (summarization, Q&A) and complex multi-agent scenarios (sequential workflows, concurrent execution, group chat). It abstracts away boilerplate LLM integration code and provides patterns for common orchestration use cases like writer-reviewer loops and specialized agent teams. Installation is straightforward—either install the full framework with all integrations or pick specific sub-packages (core, foundry, copilotstudio) to keep dependencies lean.

Use it for:

• Build a single agent that summarizes documents or answers questions using OpenAI or Azure OpenAI as the backbone.
• Create a multi-agent workflow where specialized agents (e.g., writer, reviewer, editor) collaborate sequentially or in parallel to refine content.
• Attach custom Python functions as tools to agents so they can call weather APIs, database queries, or other services during execution.
• Orchestrate agent handoff patterns where one agent routes tasks to another based on expertise or context.
• Integrate with Microsoft Foundry or Copilot Studio for enterprise LLM and workflow management.

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Builds and orchestrates AI agents powered by OpenAI and Azure OpenAI APIs, with support for multi-agent workflows, tool calling, and agent-to-agent collaboration patterns.

Yes. The package is actively maintained, production-stable, has no known vulnerabilities, and offers low install friction. It's a solid choice if you're building AI agents on OpenAI or Azure OpenAI and want structured orchestration patterns without writing boilerplate. The permissive MIT license and broad Python version support (3.10–3.14) add to its appeal. Start with a single agent to validate your use case, then scale to multi-agent workflows as needed.

Install

agent-framework on PyPI

pip install agent-framework

uv add agent-framework

poetry add agent-framework

Installing agent-framework

Before you install

Active maintenance with recent release; low install friction as a pure Python wheel with a single core dependency. Supports Python 3.10 through 3.14 and shows strong adoption (top 5000 PyPI by downloads).

License in practice

MIT license (permissive) allows unrestricted use, modification, and distribution in commercial and private projects.

Quickstart

pip install agent-framework

import asyncio
from agent_framework import Agent
from agent_framework.openai import OpenAIChatClient

async def main():
    agent = Agent(
        client=OpenAIChatClient(),
        instructions="You are a helpful assistant."
    )
    result = await agent.run("Your prompt here")
    print(result)

asyncio.run(main())

Requires Python 3.10+. Requires OpenAI or Azure OpenAI API credentials set as environment variables (OPENAI_API_KEY or AZURE_OPENAI_API_KEY, etc.).

Verify before relying

• Whether agent-framework-core and integration packages (foundry, copilotstudio) are included in the main agent-framework install or require separate installation.
• Specific performance characteristics or limits for multi-agent orchestration (e.g., concurrent agent scaling, latency).
• Whether the framework supports non-OpenAI LLM providers beyond Azure OpenAI.

Package facts

Evidence: agent_framework-1.14.0-py3-none-any.whl

Tags

More Artificial Intelligence packages

Further reading