Mistral's Shift to MCP for Data Retrieval Requires Manual Migration and Raises Data Governance Questions

Transition to MCP for Data Retrieval

Mistral is changing how its Vibe Work platform accesses company documents from Google Drive and SharePoint. The current system involves Mistral processing files and storing a searchable index in its European data centers. The new approach utilizes MCP (a common interface for models to call tools and retrieve data from external services) to handle these connections.

Manual Migration Required

The transition to MCP is not automatic. Administrators will need to manually install the MCP replacements before users can reconnect their Google or Microsoft accounts. This manual step is necessary to ensure continued access to company documents through Vibe Work.

Shift in Retrieval Architecture

Under the previous system, Mistral managed the retrieval process by searching its prebuilt index, with permissions checked against copies from Google Drive or SharePoint. The new MCP-based system shifts retrieval off-platform, meaning third-party connectors will handle how documents are retrieved. Mistral has not provided details on the retrieval architecture behind these new connectors or who operates the underlying MCP servers.

Unclear Data Governance and Permissions

The move to third-party operated MCP servers for data retrieval raises questions about data governance and how customer data will be handled. Mistral states it does not run these third-party servers, and therefore cannot guarantee their behavior or how they will manage customer data. This creates a gap in the governance Mistral previously provided for indexed data.