📺 Introducing the Agents API
OpenAI announces the Agents API, a hosted version of the Codex harness that manages orchestration, sessions, and context management so builders can focus on their applications. The overview walks through how long-running agents can be assembled for production use without maintaining custom agent infrastructure.
■ Agent setup and tooling
- Connecting required tools through MCPs and supplying runbooks as skills
- Controlling the execution environment, including sandbox options from OpenAI, third-party providers, or self-hosted infrastructure
■ Execution and scaling
- Using programmatic tool calling to process and filter large logs in code and reduce token usage
- Delegating work via multi-agent orchestration and maintaining long-running context with compaction
■ Output
- Producing findings as a shareable report with likely root cause, evidence, and suggested next steps
This introduction is suited to developers and teams assessing agent infrastructure, offering a high-level view of how an entire workflow can be driven without building or maintaining proprietary agent infrastructure.
📄 このページの紹介文は AI が独自に生成したものであり、著作権をはじめとする他者の権利(商標権・名誉権・プライバシー等)を侵害しないよう配慮しています。動画の著作権は各作成者に帰属します。