Box Architect brings autonomous agent workflows to Box

Box Architect builds on Box's existing content management and security foundation, extending it to autonomous agent workflows. The announced features—workflow design in plain language, testing with real content before production, and deployment with monitoring—are aimed at letting teams move from pilot to production with confidence. The tool's ability to select from multiple major model providers at different workflow stages gives teams flexibility on cost and capability.

Security is a central theme. Box says content stays within Box throughout workflows, and existing access permissions, classifications and retention policies apply at every step. The agent security controls currently available—classification-based policies, MCP guardrails, and threshold-based alerts—are complemented by planned threat protection features for early 2027, which are designed to detect and contain attacks such as mass data exfiltration or account takeover. These additions suggest Box is preparing for a future where agents act on corporate content more autonomously, and where managing what they can access and do becomes a core part of the platform's value.