Box is pleased to announce Box Context Retrieval, a managed, permissions-aware context engine that allows developers to ground enterprise AI agents in Box content without moving data or maintaining custom vector databases. Context Retrieval uses hybrid semantic and keyword search to extract and rank relevant document passages directly where your enterprise knowledge is already stored and governed, reducing token costs and ensuring accurate AI responses.
- Connect your AI agents or agentic workflows to Box using the Context Retrieval API or MCP integration.
- Query across your enterprise content or scope retrieval to specific Box Hubs and folders for curated context.
- Receive focused, ranked text passages with source file attribution and live permission enforcement.
Before passing retrieved content to downstream models, Box automatically verifies live user permissions and Shield classification policies, ensuring agents only see information the requester is authorized to view.
Box Context Retrieval is available via API and MCP to developers and enterprise accounts, requiring no external vector database or data synchronization configuration.