A new hosted Codex runtime enables the creation of cloud-based AI agents that can operate in isolated computing environments via OpenAI's Agents API. Currently in public beta, the interface manages the agent loop, coordinating model calls, tool use, and context management to handle long-running sessions. Developers define the specific model and tools the agent employs, while the system can deploy multiple sub-agents to perform tasks in parallel.
The infrastructure allows agents to run code, install packages, and produce artifacts within Linux sandboxes. OpenAI provides these hosted environments directly or offers 1st party integrations with external providers including DigitalOcean, Daytona, E2B, Oracle Cloud, and Vercel. This API takes over agent orchestration and state management that developers previously had to build from scratch.
Key sources
- SOURCE@openaidevs“OpenAI runs the agent. Daytona gives it a computer.”x.com
- SOURCE@openaidevs“support a 1st party integration for @OpenAI’s Agents API”x.com
- SUPPORT@openaidevs“OpenAI's new Agents API is a hosted Codex runtime that can connect to sandboxes like E2B”x.com
- SUPPORT@reach_vb“What does the new Agents API mean for the apps you use, and how does it fit alongside OpenAI’s other products?”x.com
- SUPPORT@maxforai“OpenAI 直接把 Codex 背后那套 Agent Harness，做成 API 卖了”x.com
- SOURCEhuggingnewshuggingnews.com