OpenAI — Agentic AI Overview
Overview
OpenAI's contributions to agentic AI include foundational design pattern documentation, the AutoGPT reference implementation, and the Responses API with built-in tool use. Their agentic design patterns paper is one of the most cited references for structuring multi-step agent workflows.
Key Offerings
| Product / Area | One-liner | Wiki Reference |
|---|---|---|
| OpenAI Codex | Cloud coding agent — isolated sandbox, parallel tasks, Skills, Automations, computer-use via GPT-5.4, Codex Security | AgenticFrameworks/ai-coding-agents.md |
| Agentic Design Patterns | OpenAI's canonical taxonomy of agent workflow patterns (reflection, tool use, planning, multi-agent) | DesignPatterns/openai-patterns.md |
| Practical Guide to Building Agents | End-to-end production guide covering agent definition, model/tool/instruction design, orchestration (single-agent, manager, decentralized), guardrails, and HITL — distilled from OpenAI customer deployments | DesignPatterns/openai-patterns.md |
| AutoGPT | Early open-source autonomous agent demonstrating long-horizon goal decomposition | AgenticFrameworks/autogpt.md |
| OpenAI Skills Repository | Official GitHub repository of ready-made skills for OpenAI Codex and Responses API-based agents | Standards/skills.md#provider-skills-repositories |
| Loop Engineering (Codex app) | Automations tab (project, prompt, cadence, environment) with a Triage inbox, built-in per-thread worktrees, TOML subagents in .codex/agents/, and /goal run-until-done — composed into self-feeding automation loops |
AgentHarness/loop-engineering.md |