OpenAI Opens the Agents API, the Infrastructure Behind Codex and ChatGPT
OpenAI introduced the Agents API, a public-beta platform that exposes the same infrastructure and management layer powering Codex and ChatGPT’s own long-running agents to outside developers.
Details
- One-call setup: developers create a production-ready agent by specifying the task, model, tools, and environment in a single API call
- Environment choice: agents can run in OpenAI-managed sandboxes, self-hosted infrastructure, or partner-provided environments including E2B, Vercel, and DigitalOcean
- Context management: the system automatically compacts earlier conversation history as a session approaches its token limit, letting workflows span multiple context windows without manual handling
- Tool efficiency: intelligent tool search loads only relevant tool definitions to cut token costs, while supporting parallel tool execution and chaining
- Multi-agent orchestration: the API can break a complex task into independent pieces and delegate them to subagents running in parallel, each keeping its own context while a primary agent coordinates the results
- Built on Codex harness: the API is powered by the open-source Codex harness, whose codebase is publicly available on GitHub
- Pricing: public beta with no extra platform fee — usage is billed at standard OpenAI rates for the tokens and tools an agent actually consumes
What happened next
By exposing Codex’s own agent orchestration layer as a general-purpose API, OpenAI is positioning long-running, multi-agent workflows as a standard building block for third-party developers rather than something unique to its first-party products — arriving the same week as GPT-Live-1 and ChatGPT for Financial Services as part of a broader developer-platform push.