OpenAI Opens Codex-Powered Agents API to Developers

OpenAI has launched the Agents API in public beta, giving developers access to the managed harness and infrastructure behind Codex. OpenAI News says the API is designed for agents that need to preserve context, use tools, coordinate subagents, and keep working for hours or days.

A single API call can define an agent’s task, model, tools, environment, and input. Developers can connect Model Context Protocol servers, enable multiple subagents, mount knowledge vaults, and direct agents to save files and results. OpenAI maintains the harness, while teams can choose an OpenAI-hosted sandbox, their own infrastructure, or a partner environment.

Infrastructure choices

The launch includes integrations with Blaxel, Cloudflare, Daytona, DigitalOcean, E2B, Modal, Oracle, Runloop, and Vercel. Those options are aimed at different requirements for compute, storage, secrets, networking, cold starts, and deployment inside a company’s virtual private cloud.

For AI builders, the main shift is that orchestration and runtime reliability no longer have to be built from scratch. The application can focus on domain tools, data, and workflow design while the API handles sessions, tool execution, and multi-step agent behavior.

Early customer reports point to measurable gains, though they come from individual deployments. Ciridae reported an evaluation increase from 0.71 to 0.85 and a fourfold latency reduction. SafetyKit said its cost per case fell 60%, while Hypha reported 86% fewer failed agent responses after separating the harness from the sandbox. Nash.ai described production agents handling logistics workflows that can run for hours or days.

Source: OpenAI News