Protocols & Patterns

AI Agents

An agent is an LLM in a loop, with tools and memory. The model reasons about what to do next, tools let it act, and memory lets it carry context. This page covers the patterns, what each vendor ships (Cowork, ChatGPT Agent Mode, AgentCore), and what it takes to put an agent in front of users.

In this topic

Everything about AI agents and agentic systems

Reference

A complete walk through agents: plan-act-observe loop, ReAct and Reflexion, tool use with MCP, the four memory types, planning and reasoning models, multi-agent orchestration patterns (orchestrator-as-tools, handoffs, supervisor-and-graph), computer use, the vendor product landscape (Anthropic Cowork, ChatGPT Agent Mode, Microsoft Copilot Cowork + Agent 365, Gemini Enterprise, AWS AgentCore), evaluation (SWE-Bench, Terminal-Bench, OSWorld), production patterns (sandboxing, dispatch, observability, cost), identity (Entra Agent ID, AgentCore Identity, OAuth on-behalf-of, composite identity), governance (EU AI Act, NIST AI RMF, ISO 42001), frameworks compared (LangGraph, CrewAI, OpenAI Agents SDK, Claude Agent SDK, Strands, ADK), the agentic computing paradigm shift, and a hands-on six-step build.

Last updated May 16, 2026