OpenAI has released its Agents API in public beta, giving outside developers direct access to the managed agent infrastructure that previously ran only Codex and ChatGPT. Both The Decoder and InfoWorld reported the release on September 11, 2026.
According to InfoWorld, the API bundles orchestration, context management and execution infrastructure, and can help developers design a custom agent in a single API call after they specify the task, model, tools and environment. The Decoder reported that built-in capabilities include code execution, file processing, parallel tool use and delegation to sub-agents, and that the service is built on the open-source Codex harness and supports the Model Context Protocol, custom functions and integrated tools such as web search. Agents can run for hours at a stretch.
Choice of execution environment
For running agent code, developers can pick an OpenAI-managed sandbox, their own infrastructure, or a partner environment. InfoWorld lists Blaxel, Cloudflare, Daytona, DigitalOcean, E2B, Modal, Oracle, Runloop and Vercel among the supported sandbox providers. There is no separate platform fee: The Decoder reported no additional charges beyond token usage.
InfoWorld reported that startups, SaaS companies and enterprises already using OpenAI are the most likely early adopters because the API reduces development effort, while enterprises pursuing multi-model strategies may prefer to keep their own harness or take a hybrid approach.
Lock-in and data concerns flagged
Analysts quoted by InfoWorld said the release removes much of the engineering overhead of running agents at scale. Pareekh Jain, principal analyst at Pareekh Consulting, said "the main advantage with the Agents API is fewer moving parts," noting that a hand-built long-running agent needs a job queue, a state database, a sandbox fleet, a compaction routine and a retry policy, each of which someone has to own. Phil Fersht, chief executive of HFS Research, said enterprises "may need fewer engineers to build the infrastructure around each agent, which should reduce development time." Amit Kumar Jena, AI development head at Kanerika, said that reduction in engineering work could help chief information officers cut the time needed to put an agent into production and then scale it.
Jain also called lock-in the biggest concern, arguing that if OpenAI supplies the model, context management, tools, orchestration and execution environment, moving to another platform becomes harder; InfoWorld reported that such dependency could weaken an enterprise's negotiating position on pricing and terms. The outlet added that the API does not support Zero Data Retention even when an enterprise uses its own sandbox, and named Anthropic's Claude Managed Agents, AWS Bedrock AgentCore, Microsoft's Foundry Agent Service and the open-source LangGraph framework as existing alternatives. The Agents API remains in public beta, with pricing tied to existing token and tool-call rates rather than a separate subscription tier.