OpenAI's Agents API enters public beta as Box Mount links enterprise content to the sandbox

OpenAI has announced that its Agents API is now in public beta. The API gives developers a fully managed, hosted sandbox in which agents can run tools, coordinate work, and handle complex tasks. According to OpenAI, the API is built on the Codex harness and handles orchestration, long-running sessions, and context management, allowing developers to focus on what makes their agents unique.

Box also introduced Box Mount, which brings enterprise content into the same sandbox as files that agents can read, reason across, and use to produce new work. Box Mount works with normal shell commands and file paths, and handles two-way synchronization automatically. In a demo, a lead agent mounted a deal room from Box, read five source files, launched three specialist agents in parallel, and wrote four reports back through Box Mount with an initial verdict. When a customer revised an MSA in Box, Box Mount synced the updated version into the sandbox, prompting agents to reassess and refresh all four reports while preserving the revised MSA as version two in Box.

Box Mount is currently in private preview. The result is a shared workspace where humans and agents work on the same governed Box content without custom file-transfer logic, with permissions, versions, and audit history preserved throughout.