How Mentornity AI and the MCP connector use programme data, and what they never do with it.
01Mentornity AI
- The AI assistant is switched on per programme, at the organisation’s request.
- It runs on Anthropic’s Claude models, either on Mentornity’s account with prepaid credits or through the organisation’s own Anthropic API key. With its own key, the organisation also holds the commercial relationship and its own data controls with the model provider.
- Anthropic does not train its models on data sent through its commercial API.
- Programmes whose matching uses the AI rule send the text of participant profiles to TypeSafe, which returns match scores; programmes that do not use the rule send nothing.
- The assistant acts with the permissions of the person using it: it can read and change only what that person could in the web application.
02MCP connector
Members can connect their own AI assistant to Mentornity through the Model Context Protocol (MCP) connector. Access is granted by the member through an OAuth sign-in, is limited to an allow-list of tools, can be switched off per programme, and acts exactly with that member’s permissions. Every action that changes something is shown to the member as a preview and runs only after they confirm it. Text written by other people is marked as untrusted for the assistant, as a guard against prompt injection.
03Contact
Questions about this policy, security questionnaires and Data Processing Agreement requests: privacy@mentornity.com.