> ## Documentation Index
> Fetch the complete documentation index at: https://docs.mobbin.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Privacy & data

How Mobbin handles the data that flows through the MCP server, Standard Search and Deep Search. For
the full picture, see our [Privacy Policy](https://mobbin.com/privacy) and
[Trust Center](https://trust.mobbin.com).

## What data is stored from MCP requests?

MCP requests are recorded in our product analytics with the tool called, its parameters (including
query text), the timestamp, and the user and workspace identifiers of the connected account. This
data is retained for as long as needed for product analytics and service operation, in line with our
[Privacy Policy](https://mobbin.com/privacy).

## Are queries used for AI training?

Mobbin does not currently train or fine-tune AI models on customer search queries. AI query
interpretation is part of how Standard Search and Deep Search process a query, so there is no
separate opt-out. Requests to our AI query-interpretation providers are sent with prompt training
disabled.

## Which AI providers process search queries?

Standard Search and Deep Search queries are processed through Vercel's AI Gateway and Google Cloud.
Both are listed at [trust.mobbin.com/subprocessors](https://trust.mobbin.com/subprocessors). No
user, account or workspace identifiers are included in the query text sent to these providers.

## What can the MCP server access?

Each connection is authorized by an individual user and tied to one Mobbin workspace. The server can
access Mobbin's design library, the connected workspace's collections, and public collections. It
cannot reach other workspaces' private collections or anything outside the permissions that user has
in the Mobbin web app. You can [revoke access](https://docs.mobbin.com/mcp/disconnect.md) for any client at any time.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.