Skip to main content
MCP Reviewed Queries extends the existing reviewed queries system to support visualizations generated from Model Context Protocol (MCP) tool calls. When you provide positive feedback on an MCP-generated visualization, it becomes a high-quality example that the system uses to improve future responses for similar queries. This feature creates a continuous learning loop in which your everyday interactions help the system understand what a correct answer looks like, forming patterns it can repeat. This guide will walk you through the steps for asking questions, approving results, and improving the system’s accuracy over time.

How it works

The process follows three steps to turn a single correct answer into a repeatable pattern for the system.

Ask a Question

Start by asking a question in chat that requires a connected tool to generate a response. The system will use the tool to provide data and a corresponding visualization.

Approve the Result

Inspect the visualization and the data provided. If the result is accurate and meets your requirements, click the Thumbs Up icon. This action flags the response as a validated example.

Get Better Answers

The system automatically uses that approved example to guide its logic when answering similar questions in the future. Over time, these reviewed queries reduce inconsistency and ensure the system follows your preferred patterns.

Key capabilities

MCP Reviewed Queries allow you to improve system performance without complex manual setup:
  • Automatic Improvements: Enhance tool-based answers simply by interacting with the chat.
  • Reusable Examples: Turn one-off correct visualizations into reference points for the AI.
  • Pattern Recognition: Help the system learn real-world usage patterns that metadata and schemas alone cannot capture.
  • Consistency: Reduce variations in how the system responds to similar questions across your organization.

Target users

The MCP Reviewed Queries feature is available to Admins and Explorers who rely on validated query patterns to ensure accuracy, reduce inconsistency, and maintain high-quality data standards across their organization.

FAQs

The system treats that specific result as a high-quality example. It uses the logic and output of that interaction to guide its handling of similar requests in the future.
No. The learning process happens automatically once you approve a result. You do not need to perform additional configuration in the Domain settings.
Yes. Approved examples help the system respond more effectively to similar questions for all users within the Domain, while strictly respecting individual access controls and permissions.
Basic feedback and approval are supported currently. More advanced visibility and management options for these examples are planned for future updates.

Next steps