Dropbox has integrated Model Context Protocol with Dash, its internal knowledge platform, to bring security design context into AI-assisted code review. The system is meant to help reviewers compare a pull request with the threat models and security requirements that shaped the feature.
Model Context Protocol, or MCP, is a way for AI tools to retrieve external context through structured connections. In Dropbox’s setup, that context can include design documents and security expectations that may not be visible from the changed code alone.
The practical problem is common in large engineering teams. A reviewer can see what code changed, but not always why a security control was required or which risk it was supposed to address. Surfacing that background during review can make the AI assistant more useful and reduce the gap between design intent and implementation.
InfoQ’s report frames the work as an internal architecture lesson, not a packaged product announcement. The important limitation is that the value depends on the quality and freshness of the documents the system retrieves.