Table of Contents
Key Insights
Unwrap's MCP
Unwrap maintains a production MCP (Model Context Protocol) server that exposes the same toolset that powers the in-app Unwrap Assistant, so any compatible AI platform (Claude Desktop, Claude Web, Claude Code, ChatGPT, Cursor, and others) can natively query an Unwrap workspace in plain English. Step-by-step setup instructions for each platform live in the Unwrap MCP knowledge base article.
What Is an MCP, and Why Does It Matter for Customer Feedback?
MCP, short for Model Context Protocol, is an open standard developed by Anthropic for connecting AI models to external data sources and tools. Instead of copying customer feedback into an AI tool by hand or building a one-off integration, an MCP lets the model query the source system directly with the same auth and permissions the user already has.
For customer feedback, that matters because the value lives in being able to ask the question that just came up: "what are users saying about the new checkout flow this week?", "which features are driving NPS detractor verbatims?", "is the housekeeping complaint volume up at our Austin properties?" Until recently, that kind of question required either a dashboard built in advance or an analyst pulling a report. With an MCP, the answer comes back inline in the AI platform the team is already working in.
What Can You Do with the Unwrap MCP?
The Unwrap MCP gives an AI platform access to twelve tools spanning the full feedback intelligence stack: listing views, searching taxonomy groups and segments, surfacing the most actionable issues, sampling feedback entries, generating charts over time, and running custom classification or analysis instructions across a saved sample. In practice, that translates to prompts like:
- What are my top issues for the last 90 days?
- Filter to just complaints about feature X, and show me how they've trended over six months.
- Find me customer feedback around feature Y, and write a priority-ranked PRD with customer anecdotes and links.
Because the model is querying real data and not a static export, the answer reflects whatever showed up in the feedback pipeline that morning.
Which AI Platforms Does the Unwrap MCP Support?
The Unwrap MCP is built on the open MCP standard, so it works with any MCP-compatible platform. The docs include step-by-step setup instructions for the four most common configurations:
- Claude Desktop and Claude Web (via Organization Settings → Connectors)
- Claude Code (claude mcp add command)
- ChatGPT (via developer mode and Apps)
- Cursor (via mcp.json)
Other MCP-compatible platforms can connect to the same endpoint as long as they support OAuth.
How Do You Connect the Unwrap MCP?
Connection is a one-time OAuth flow. In Claude Desktop, for example, an admin adds the Unwrap MCP as a connector under Organization Settings, clicks Connect, completes the redirect, and the MCP is live for the workspace. The full instructions for each supported platform live in the Unwrap MCP docs, including a static OAuth option for orgs that want one set of shared credentials across all users.
Is the Unwrap MCP Read-Only?
Yes. All twelve tools exposed by the Unwrap MCP are read-only. The connected AI platform can query, search, sample, analyze, and visualize feedback data, but it cannot create, edit, or delete anything in the Unwrap workspace. Access is scoped to the views the authenticated user already has permission to see.
Who Is the Unwrap MCP For?
Anyone whose work depends on customer feedback, but whose primary working surface is increasingly an AI platform rather than a dashboard. For product managers running discovery, support leaders triaging emerging issues, CX teams drafting executive briefs, and engineers tying feedback to specific feature work, the MCP collapses the round-trip between "I have a question about what customers are saying" and "I have an answer" into a single chat turn, in the tool the user was already in.
For the full setup walkthrough and the current tool reference, see the Unwrap MCP documentation.
Frequently Asked Questions
