Box

Box Mount is a developer tool from Box that gives sandboxed AI agents a working file system backed by a Box folder. A developer provisions an agent sandbox, brings Box content into that environment, and returns the agent's work to Box. The launch is aimed at teams building agent workflows in runtimes like E2B, where the pain of writing custom download, upload, retry, and conflict logic around every document tends to eat more time than the agent itself. Under the hood, Box Mount gives an agent sandbox a POSIX-compatible Box workspace, allowing code and tools inside the sandbox to work with Box content through familiar file paths , with synchronization running in both directions so a person editing a file in Box and an agent writing to the mounted path stay in the same working copy. Because reads and writes flow through the Box API, Box continues to provide the content layer, including permissions, versions, classifications, retention, Shield protections, and audit history , which matters for teams that cannot ship agents that bypass enterprise governance. The design is framework agnostic, so an OpenAI agent, an Anthropic agent, LangGraph, CrewAI, a Python script, or a CLI tool can operate against the same mounted directory. The launch ships with a reference implementation, a contract-review app that mounts a Box folder into an E2B sandbox, runs an agent against an MSA and an approved playbook, writes a review memo back to Box, and assigns a task to a human reviewer. Box Mount is currently in private preview, and the demo and accompanying write-up were put together by Andrew Wong, a developer advocate at Box, for developers exploring document-heavy agent patterns like RFP drafting, deal-room analysis, and compliance review.