1. Connect your client
- Activate Pro and open MCP settings.
- Enable MCP and choose the dataset folders to allow.
- Copy the generated configuration into your AI client’s MCP settings.
- Reconnect the client.
2. Review the available tools
Open the tool list in RowCove, inspect a tool’s inputs and run a test call. Call history helps you review what was requested.
3. Start with one question
For example: “Inspect the schema of this dataset, then show records from one date partition. Read only the fields needed and leave the files unchanged.”
Review the result and check a representative record in the desktop inspector before expanding the scope.
4. Share deliberately
Requested results go to your connected AI client. If it uses a cloud model, those results may reach its provider. Allow only the folders and fields needed for your task.
Troubleshooting
If the connection fails, check that Pro and MCP are active, copy the current configuration again and reconnect. Check the folder permissions and call history for the reported error.
Contact support without sending dataset contents, passwords or access tokens.