OpenAI has launched the public beta of the "Agents API" for developers. By using this API, developers can create production-ready agents through a single API call simply by specifying the task, model, tools, and environment. There is no additional fee for using the API; only the standard charges for the tokens and tools used by the agent will apply.

For the execution environment, users can choose between an OpenAI-managed sandbox, their own infrastructure, or one of OpenAI's sandbox partners. The OpenAI-hosted sandbox leverages the same infrastructure as Codex and ChatGPT, providing a secure environment for agents to execute code, manipulate files, and create deliverables.

Additionally, the API includes context management features to ensure agents retain information during long sessions. This allows for the continuation of processes by preserving previous context when a session approaches its limits. Furthermore, it supports multi-agent configurations, enabling complex tasks to be split among multiple sub-agents operating in parallel, which is expected to accelerate tasks such as research, analysis, and coding.

Source: