OpenAI Docs

Agents

• Updated the Agents guide to compare the Agents API, Agents SDK, and Responses API and point you to the right docs for your app.
• Added Agents SDK for how the SDK fits together and where to go next.

Agents API

New guide set for building durable cloud agents with a managed Codex harness:

• Agents API — overview
• Agents API quickstart — create an agent that writes and runs a script in an OpenAI-hosted sandbox
• Architecture — harness, environment, and application server
• Configuring Agents — define an agent, reuse config, and customize sessions
• Multi-agent — delegate work to independent subagents
• Observability and usage — live progress, completed work, and token usage
• Tracing — inspect turns, model responses, and tool calls

Sessions

• Run and continue sessions
• Events and items
• Manage sessions
• Session webhooks

Environments and sandboxes

• OpenAI-hosted sandboxes
• Self-hosted sandboxes
• Sandbox lifecycle
• Files and artifacts
• Sandbox security

Provider guides:

• Blaxel
• Cloudflare
• Daytona
• DigitalOcean
• E2B
• Modal
• Oracle Cloud Infrastructure (OCI)
• Runloop
• Vercel

Tools

• Functions
• MCP connections
• Plugins
• Vaults
• Web search

Skills

• Updated the Skills description to cover reusable skills for Responses API shell tools and Agents API sandboxes.