**[OpenAI Opens Codex Agent Core Capabilities to Developers, Agents API Enters Public Beta]** Beating AI News Flash: OpenAI has officially launched the public beta version of the Agents API, granting all developers access to the agent framework and infrastructure previously used to support Codex and the enterprise version of ChatGPT. Developers can directly build production-grade AI agents via the API, with fees calculated based on token consumption and tool usage. The Agents API focuses on long-duration operations and complex task execution capabilities. Developers only need to specify the task, model, tools, and runtime environment in a single API call to create an agent. Its core features include automatic context compression, parallel tool invocation, and multi-agent collaboration, enabling agents to run continuously for hours or even days. Customer cases disclosed by OpenAI show significant improvements: SafetyKit reduced single-case processing costs by 60%, Hypha lowered agent response failure rates by 86%, and Cirridae improved evaluation scores from 0.71 to 0.85 while reducing latency by a factor of four. In terms of infrastructure, the Agents API can utilize the same sandbox environment as Codex and ChatGPT. It has also integrated with partners such as Blaxel, Cloudflare, DigitalOcean, Oracle, and Vercel, supporting various CPU, GPU, memory, and deployment environments. OpenAI further stated that the Agents API is built on the open-source Codex execution framework and supports tools like MCP, custom functions, and web search. This launch signifies OpenAI's expansion of Codex from an AI programming product into a developer-oriented agent infrastructure platform. [Original Link]