Brought to your by Mark.

Open AI SDK

platform.openai.com/docs/guides/agents

New agent sdk from open ai including tools such as computer use, function calling, file search, web search

pip install openai-agents
from agents import Agent, Runner

agent = Agent(name="Assistant", instructions="You are a helpful assistant")

result = Runner.run_sync(agent, "Write a haiku about recursion in programming.")
print(result.final_output)

# Code within the code,
# Functions calling themselves,
# Infinite loop's dance.

Overview

OpenAI Agent SDK is a lightweight yet powerful framework designed for building multi-agent workflows. The framework enables developers to create, coordinate, and monitor AI agents with a focus on simplicity and control. It’s compatible with any model providers that support the OpenAI Chat Completions API format.

Features

  • Core Agent Components

    • Agent Configuration with Instructions
    • Tool Integration Support
    • Configurable Guardrails
    • Agent-to-Agent Handoffs

  • Tracing & Monitoring

    • Built-in Run Tracking
    • Debugging Capabilities
    • Workflow Optimization Tools
    • Support for External Destinations
      • Logfire Integration
      • AgentOps Integration
      • Braintrust Integration

  • Safety & Control

    • Input Validation
    • Output Validation
    • Configurable Safety Checks
    • Handoff Control Mechanisms

Design Principles

  1. Lightweight Architecture

    • Minimal dependencies
    • Simple API design
    • Focus on essential functionality

  2. Extensibility

    • Custom span support
    • Pluggable tracing system
    • Flexible tool integration

  3. Developer Experience

    • Easy setup process
    • Clear documentation
    • Comprehensive examples
    • Built-in debugging tools

  4. Interoperability

    • Compatible with OpenAI Chat Completions API
    • Support for multiple model providers
    • Integration with existing tools and frameworks

Built with OpenAI Agent SDK

Examples of applications and systems built using the framework:

  1. Multi-Agent Systems

    • Coordinated agent workflows
    • Complex task delegation
    • Agent communication networks

  2. Monitored Applications

    • Traced agent interactions
    • Debuggable workflows
    • Performance optimization systems

  3. Safety-First Applications

    • Input/output validated systems
    • Controlled agent handoffs
    • Guardrail-protected workflows

Authors

Core Contributors:

  • @rm-openai
  • @dmitry-openai
  • @jhills20
  • @stevenheidel
  • @pierDipi
  • @chuang-openai

Source 1 Source 2

Note: The OpenAI Agent SDK is actively maintained and developed as an open-source framework, encouraging community contributions while maintaining high standards for code quality and security.