OpenAI has officially launched its Agents API in public beta, allowing developers to build cloud-based AI agents that can run for hours, execute code, and process files. Powered by the same infrastructure behind Codex and ChatGPT, the new API introduces features like automatic context management, parallel tool usage, and task delegation across sub-agents.
Developers using the API can choose between OpenAI-hosted sandboxes or third-party partner environments including Cloudflare, Vercel, and Oracle. Built upon the open-source Codex harness, the API supports MCP, custom functions, and native tools such as web search.
OpenAI stated that there are no additional platform fees for using the Agents API, with billing calculated solely on token usage.
Why it matters
Enables long-running, autonomous agent workflows natively on OpenAI’s core infrastructure without added platform overhead.
Simplifies complex multi-agent architectures via built-in sub-agent delegation and context management.
Provides flexibility in execution environments through support for partner sandboxes like Vercel and Cloudflare.
Source: the-decoder.com
