OpenAI launched the Agents API in public beta for all developers. It exposes the same harness and infrastructure that runs Codex. Developers specify a task, model, tools and environment in one API call. Agents can run for hours or days across multiple context windows. The service automatically compresses earlier context as limits approach. Tool search loads only the tool definitions an agent needs. Agents can also split work across subagents that run in parallel. Compute can run in an OpenAI-hosted sandbox, on a developer's own servers, or with partners such as Cloudflare, Vercel, Modal, Oracle and DigitalOcean. The harness is open source, so its logic can be inspected. There are no extra fees beyond tokens and tools used.
What changed
Developers had to build and maintain their own agent orchestration, context handling and sandbox infrastructure.
What it unlocks
Creating a long-running cloud agent in one API call, using OpenAI's managed Codex harness and a chosen sandbox.
- no extra fees, tokens and tools only
- 9 sandbox partners
- 3 max concurrent subagents in example
What you need to act on it
- an OpenAI API account
- public beta access
Sources