The Retently MCP gives your AI agent a set of tools for reading and acting on your Retently data. Once connected, you can ask questions and request actions in plain language, and the agent picks the right tools to carry them out.

The MCP works the same way across any AI agent you connect it to, such as Claude or OpenAI Codex. The examples below use plain prompts that apply to all of them.

This article explains the tool categories and shows example prompts.

Many of the capabilities below depend on the Personal Identifiable Information Access toggles on your AI Connections page. Filters, contact properties, customer attribution, and every write action require these to be enabled. With them off, your AI agent is limited to anonymous, quantitative data. Enable them before you connect, and if you already connected with them off, reconnect to apply the change. Unless you need to keep your data anonymous, we recommend enabling both. See any connect article for details.

The best way to understand what is available is to ask your AI agent directly:

What can the Retently MCP do? List all the tools and give a short description of each.

Your AI agent returns the tools grouped into categories. This gives you a map of what you can ask for before you start. The MCP includes more than 50 tools, so you do not need to memorize them or name a specific tool in your requests. The agent reads the list and picks the right tools for each question.

Want to see the whole surface area first? The Retently MCP Tools and Glossary article lists every tool with a short description, plus a glossary of Retently terms. Skim it to analyze the tools on your own before trying the examples below.

A single request often uses several tools at once. Some tools read information (for example, looking up another user's ID), and other tools use that information to perform an action (for example, assigning a response to that user). You do not need to know which tools are involved. Your AI agent chooses and chains them for you.

Read and act on survey responses. In Retently, "responses" and "feedback" mean the same thing.

Work with your survey campaigns.

Triggering a survey from a campaign is one of the more powerful actions: you can ask your AI agent to send a survey to a contact through a specific campaign without leaving the tool. This requires write access (see the note above).

Look up and manage the people in your account. "Customers" and "contacts" mean the same thing.

Get the same numbers you see in Retently, plus custom views.

A score in Retently is never a plain average of the ratings. NPS, CSAT, and CES each have their own formula. When you ask for a score, your AI agent uses the scoring tool so the number matches what you see in your account.

Your AI agent may also check your templates when you ask why a campaign's response rate is low or how a survey could be improved, since the questions live in the template.

These are some of the things you can ask once Retently is connected. Responses vary each time, since they reflect your live data.

What trends do I have in Retently?

Your AI agent identifies the trends tool and lists your trends, such as "NPS by Department" and "NPS by Product."

Build a visual report for the NPS by Department trend.

Your AI agent pulls the trend data and assembles a custom report you can open and share. This is useful when you want to visualize your data in a different way than the built-in views, or build a custom report on top of what Retently already gives you in the app and email digests.

Keep requests focused. A very broad reporting prompt can take longer and produce more than you need. Start specific, then expand.

Pull all survey responses we received today.

Your AI agent returns the responses from today with their scores and comments, matching what you would see by applying a "today" date filter on the Feedback page, without logging in.

Give me a summary of all survey responses with the Pricing topic and negative sentiment.

In Retently you would open the Feedback page and apply a topic and sentiment filter to see those responses. With your AI agent, you get the same set plus a written summary of what they say, so you skip the manual reading.

How many responses did we receive this month so far, and how many of those are from detractors who mentioned pricing?

Your AI agent pulls the totals, identifies the detractors, and checks which of those responses mention pricing, then returns the breakdown: total responses, how many were detractors, and how many of those detractors raised pricing.

When you ask about a topic, your AI agent can account for related terms, not just one exact keyword. A literal search in the app for "pricing" matches only that word, so it misses a detractor who wrote "too expensive" or "the cost." The agent can reason about those alternatives and surface responses a single-keyword search would skip.

Because your AI agent can both read your Retently data and act on it, you can set up recurring or connected workflows. A few ideas:

The MCP is still being actively improved. The best way to start is to list the tools, then try the questions you normally ask yourself when you look at your Retently data. If something is missing or could be better, let us know, ideally with the exact prompt you sent to your AI agent, so we can optimize for that scenario.