How We Brought BlockSurvey into ChatGPT and Claude with MCP

Think about the last survey you built. You probably picked a question type, typed the question, added the answer options, then went hunting for the setting that makes the next question appear only for some people. It works, but it takes a while.

Now you can just tell your AI assistant what you want, and it builds the survey in your BlockSurvey workspace. The BlockSurvey MCP works with Claude, ChatGPT, and any other AI tool that can connect to an MCP server. Here is how we got here.

It all started with one message

Yohan Tengra is the founder of Biohacking India, a functional medicine practice in Mumbai, and he has been one of our customers for a while. One day he wrote to us:

"I am building an internal AI workflow in ChatGPT Business, and one important requirement is being able to operate BlockSurvey directly from ChatGPT."

He wanted to create surveys, edit questions, add branching and publish, all without leaving ChatGPT. We read that and thought, he is probably not the only one who wants this. So we got to work.

Version one: say it, and it's built

We began with the things you do in almost every survey. With the first version, you can ask your assistant to create a survey, add or delete a question, and add or delete an answer option. You can reorder questions and mark a question as required. You can edit the welcome screen and the thank-you screen, publish the survey, and ask for the share link.

It sounds simple, and it is. But it changed what a first draft feels like. You describe the survey, and a few moments later it is sitting in your workspace. You can open it in the builder and keep editing it like any other survey, because that is what it is.

Version two: logic and scores, in plain English

The next thing we tackled was logic, because that is the part most people put off. In the second version, you can add jump logic, display logic and answer logic just by describing them. Say you want people who give a low rating to see a follow-up question, or you want some respondents to skip ahead. You say so, and the assistant sets up the rule. Thank-you screens can follow logic too, so different people can see different endings.

Then another customer, Tim Peterson, got in touch. Tim is a clinical psychologist, and he wanted to add numeric expressions, which are calculated values such as scores that can build on the answers to your questions and on each other. We added them to the MCP, and Tim tried them out and shared this testimonial with us:

Testimonial from Tim Peterson, PsyD, Clinical Psychologist

"Nice work! I was able to use Claude to easily create the calculated variables in the outcome page."
"I appreciate your prompt attention to updating MCP functionality to allow me to create the variables with Claude. Big help!"

Both versions came from the same place: a customer told us what they needed, and we built it.

Three messages, one finished survey

Here is how it might go in practice. You start with a request, then keep building on it with follow-ups, the same way you would talk to a colleague.

  • "Create a customer satisfaction survey with 8 questions. If someone rates us below 3, ask what went wrong."
  • "Add an NPS question to my customer satisfaction survey, and skip the follow-up if they score 9 or above."
  • "Publish my customer satisfaction survey and give me the link."

The first message creates the survey with a rule that shows the follow-up question only to people who rate you low. The second adds a question and the logic that goes with it. The third publishes the survey and hands you the link to share.

Everything the assistant creates follows the same rules as the BlockSurvey builder. If you ask it to delete a question that other rules depend on, it checks first instead of quietly breaking your survey. And it publishes only when you ask it to.

Now live in OpenAI's plugin directory

We have some news we are happy to share. The BlockSurvey MCP is now officially listed in OpenAI's plugin directory. If you use ChatGPT, you can find BlockSurvey there and add it in a few clicks, with no setup beyond signing in.

It is a milestone for us, and it came from customers like Yohan and Tim asking for BlockSurvey in the tools they already use. Thank you to everyone who asked, tried it early and told us what worked.

The assistant can't read your responses

Privacy is why many people choose BlockSurvey, so the connector is built around it. The assistant works on the structure of your surveys only: questions, options, logic and screens. It has no way to read what respondents send in.

When you connect, BlockSurvey asks you to sign in and shows a screen with two buttons, Allow and Deny. Nothing is shared until you click Allow. The assistant can then work only in your team workspaces, which are the shared workspaces for your team on BlockSurvey. If you ever want to stop, remove the connector in your AI tool's settings.

Connect it in two minutes

The connector address is https://mcp.blocksurvey.io/mcp. For screenshots and every step, see the full setup guide.

  1. Add the connector.
    • In Claude, open Customize, then Connectors, click Add, enter a name and the address above, keep the recommended authentication options, then click Connect.
    • In ChatGPT, find BlockSurvey in OpenAI's plugin directory and add it.
    • In any other AI tool that supports MCP, add it as a remote MCP server using the address above.
  2. Sign in to BlockSurvey.
  3. Review the access and click Allow. Then ask for something. Try the three examples above, in order.

Comparing your options? See our roundup of the best MCP survey tools, or the guide to the best survey MCP for Claude.

This is just the beginning

This is only the start. We are building more functionality into the BlockSurvey MCP, including things you will not find in the MCP of any other survey platform. Keep an eye on this space.

Like what you see? Share with a friend.


Connect BlockSurvey to your AI assistant
Set up the MCP in a few minutes, step by step.
Get started
blog author description

Hari S K

Hari is a Software Development Engineer at BlockSurvey and works on the product's AI features. He built the BlockSurvey MCP server, which lets AI assistants create and manage BlockSurvey surveys directly. He also works on AI agents and decision-making systems, including JEV, and keeps learning new technologies and putting them to work in BlockSurvey.

Scripts are blocked. This site won’t work properly. If you’re using Brave, click the Shields icon and turn off Block scripts. Otherwise disable your ad blocker for this site.