OpenAI Releases Agents API in Public Beta for Autonomous Multi-Agent Workflows
Original title:OpenAI's new Agents API gives developers the infrastructure behind Codex and ChatGPT
OpenAI has rolled out a public beta of its Agents API, making the execution layer behind Codex and ChatGPT directly accessible to developers. The framework supports long-horizon cloud agents capable of running autonomously for hours, executing code, and delegating subtasks across sub-agents without platform fees beyond token usage. Integrated execution sandboxes from Cloudflare, Vercel, and Oracle provide isolated environments, marking a steady shift toward persistent background orchestration.
Why it's worth reading
By offering the infrastructure powering its internal workflows at no extra fee beyond token consumption, OpenAI significantly lowers the barrier to building complex, long-running agentic systems.