Wednesday, November 19, 2025

OpenAI API Agents SDK

OpenAI: a-practical-guide-to-building-agents.pdf


openai/openai-agents-js: A lightweight, powerful framework for multi-agent workflows and voice agents @GitHub

openai/openai-agents-python: A lightweight, powerful framework for multi-agent workflows @GitHub


openai/agents.md: AGENTS.md — a simple, open format for guiding coding agents


Agents - OpenAI Agents SDK

The OpenAI Agents SDK enables you to build agentic AI apps in a lightweight, easy-to-use package with very few abstractions. It's a production-ready upgrade of our previous experimentation for agents, Swarm. The Agents SDK has a very small set of primitives:

  • Agents, which are LLMs equipped with instructions and tools
  • Handoffs, which allow agents to delegate to other agents for specific tasks
  • Guardrails, which enable validation of agent inputs and outputs
  • Sessions, which automatically maintains conversation history across agent runs

 Examples - OpenAI Agents SDK python

openai-agents-js/examples at main · openai/openai-agents-js JavaScript

  •  Multi-Agent Workflows: Compose and orchestrate multiple agents in a single workflow.
  •  Tool Integration: Seamlessly call tools/functions from within agent responses.
  •  Handoffs: Transfer control between agents dynamically during a run.
  •  Structured Outputs: Support for both plain text and schema-validated structured outputs.
  •  Streaming Responses: Stream agent outputs and events in real time.
  •  Tracing & Debugging: Built-in tracing for visualizing and debugging agent runs.
  •  Guardrails: Input and output validation for safety and reliability.
  •  Parallelization: Run agents or tool calls in parallel and aggregate results.
  •  Human-in-the-Loop: Integrate human approval or intervention into workflows.
  •  Realtime Voice Agents: Build realtime voice agents using WebRTC or WebSockets
  •  Local MCP Server Support: Give an Agent access to a locally running MCP server to provide tools
  •  Separate optimized browser package: Dedicated package meant to run in the browser for Realtime agents.
  •  Broader model support: Use non-OpenAI models through the Vercel AI SDK adapter
  •  Long running functions: Suspend an agent loop to execute a long-running function and revive it later Future
  •  Voice pipeline: Chain text-based agents using speech-to-text and text-to-speech into a voice agent 



OpenAI @GitHub search "Agents"




No comments: