You spend half your day in an AI chat and still switch over to the survey tool for every survey, add the questions by hand and download the answers later? That switch is what the MCP server removes.
The MCP server from empirio.ai, an online survey tool from Germany, is an open interface that you connect once to an AI agent such as ChatGPT or Claude, so the agent can create, edit, publish and analyse surveys in your account without you opening the tool yourself. The MCP server for AI agents is free. By the end you will know which jobs the agent takes over and where your plan stops it.
📌 Key points at a glance
- The empirio.ai MCP server is free on every plan.
- MCP is an open standard, not just for ChatGPT.
- The server address is https://platform.empirio.ai/mcp.
- Your plan decides whether a survey can go live.
- The agent exports completed participations only by default.
Create a survey for free
With empirio.ai you can create a modern online survey in minutes — with hosting in the EU.
- AI-built survey
- Adjust by drag & drop
- Real-time analysis
What is the empirio.ai MCP server?
The empirio.ai MCP server is an access point built on the Model Context Protocol, through which an AI agent carries out real actions in your survey account: creating surveys, editing questions, publishing, retrieving results and exporting them.
MCP stands for “Model Context Protocol” and is a shared language between AI tools and other programs. Picture a universal plug. Without that plug, ChatGPT knows nothing about your surveys and can only suggest text. With the plug, the agent reaches into your account and works there.
MCP server and AI chat: the difference
A language model on its own is a genius with words and no hands. It will write eight good questions, but it cannot put a survey online, cannot produce a participation link and cannot store a single answer. The empirio.ai MCP server closes exactly that gap and the difference shows up in the chat straight away.
- AI without MCP: “Here are eight questions, you can enter them somewhere yourself.”
- AI with the empirio.ai MCP server: “The survey is created and published, here is your participation link.”
An open standard, not just for one tool
The Model Context Protocol is an open standard and not an empirio.ai invention. For you that means the empirio.ai MCP server works with any AI client that supports MCP. ChatGPT and Claude are the best known, but any other MCP-capable program can be connected too. You are tied to no single provider and can use whichever agent you already work with. The technical description sits in the open specification for the Model Context Protocol.
What can the MCP server actually do? Create and analyse surveys with AI
Through the empirio.ai MCP server your AI agent takes over the whole life of a survey, from the first question to the presentation file with the charts. You describe your goal in plain language, the agent translates it into the matching actions in your account.
Which actions those are is clearly bounded. The table below names the jobs and the plan each one needs.
| Job | What the AI agent does | Plan |
|---|---|---|
| Create a survey | a draft from a goal description or from finished questions | free |
| Edit questions | replace, add and reorder questions | free |
| Publish | go live, withdraw, revert to the draft | free |
| Retrieve statistics | views, started and completed participations, duration | free |
| Export raw data | CSV, Excel, JSON or SPSS | Lite and up |
| Cross-tabulations and filters | compare two questions, filter answers | Plus and up |
| Export charts | PowerPoint, Word or PDF | Plus and up |
As of October 2026, subject to change.
Creating, editing and publishing surveys
There are two routes into a new survey and the agent picks one from the way you phrase the request. Describe only the goal and empirio.ai writes the questionnaire itself. Supply the questions and exactly those questions are used. Either way the first thing that exists is an unpublished draft, never a survey that is already running.
Typical instructions the empirio.ai MCP server carries out:
- Create: “Create a survey with five questions on the Net Promoter Score.”
- Edit: “Replace question 3 with an open question.”
- Publish: “Put the survey live and give me the link.”
- Duplicate: “Copy last quarter’s survey as a draft.”
- Review: “Show me all my surveys and their status.”
The status the agent reports back is draft, scheduled, live or ended. The run window is the one part of a survey that takes effect without publishing: set a start or end time through the agent and it applies immediately. If you would rather not start from nothing, point the agent at the templates for surveys and questionnaires.
Retrieving and analysing answers
On the results side you simply ask in the chat. The agent pulls participant numbers, the completion rate and the distribution per question straight from your account, summarises free-text answers and puts two questions side by side in a cross-tabulation. For a report it hands you the charts as an editable PowerPoint file.
An everyday example: you lead a team of twelve in Bristol and want to know how the last release landed internally. You write “Create an anonymous survey with six questions about the release and publish it”. You share the link with the team. Three days later you ask in the same chat which scores came out lowest. Whether an anonymous survey like this really stays anonymous depends on the questions, not on the agent.
Does the MCP server lift the limits of your plan?
The empirio.ai MCP server lifts no plan limit. The connection itself is free, but the AI agent gets exactly the rights your account already has. One difference from clicking in the editor remains: the agent finishes the draft anyway and the limit only shows up when you publish.
If a draft uses a feature the plan does not cover, empirio.ai refuses to publish it and names the reason. What counts is the plan of the person who owns the survey, not the plan of the person who merely collaborates. You can check this beforehand: the agent reads the status of the survey and tells you whether publishing is allowed.
The features it hinges on in practice:
| Feature | From which plan |
|---|---|
| Participants per month | 50 on the free plan |
| Complex logic jumps | Lite and up |
| Design options | Lite and up |
| Own colours and logo | Plus and up |
| Access restriction | Plus and up |
| Multilingual surveys | Pro and up |
| Removing provider branding | Pro and up |
As of October 2026, subject to change.
Careful
An agent that creates a multilingual survey with your own logo without asking produces a draft that cannot go live on the free plan. Have it check the plan status before publishing, or state your limit in the instruction itself: “Stay within the features of the free plan.”
Your survey costs depend mainly on two things: how many people take part and which features you need.
How to connect your AI agent to empirio.ai over MCP
You need no programming knowledge for the connection. The empirio.ai MCP server is a hosted service, so you install nothing on your own machine. You enter one address in your AI client and sign in once.
Sign-in runs over OAuth 2.1, the same mechanism you use elsewhere when you log in with an existing account. There are no keys to copy by hand.
- Create an account. Register with empirio.ai, the free plan is enough to begin with.
- Enter the server address. Add empirio.ai in your AI client as an external MCP server, the address is https://platform.empirio.ai/mcp.
- Grant the rights. Sign in the first time you connect and allow the agent access to surveys and answers. Individual rights can be left switched off.
With ChatGPT and Claude
In Claude the server is added through settings as a connector of its own, in ChatGPT through the connectors in settings. Both routes need nothing but the server address, the sign-in handles the rest. For a survey with ChatGPT the route also depends on your account tier, while for a survey with Claude you will find empirio.ai straight in the connector directory. AI providers change their interfaces fairly often, so the current steps are kept in the developer documentation.
Tip
The first time round, let the agent only describe an existing survey before you ask it to create anything. One pass tells you whether the connection works, without a half-finished draft appearing in your account.
Prefer something more technical? CLI and REST API
If you would rather not work through a chat, there are two further routes and both are free. The CLI drives surveys by command in the terminal, handy for jobs that repeat. The REST API covers what the platform can do and slots into your own tools, a CRM or a dashboard. On top of that come webhooks and an n8n-verified connector. With it you can automate surveys with n8n, for example when every new response should go straight into your CRM.
If you want to start right now, an account is all it takes: at empirio.ai you can create a survey and register the MCP server in the same go.
Create a survey for free
With empirio.ai you can create a modern online survey in minutes — with hosting in the EU.
- AI-built survey
- Adjust by drag & drop
- Real-time analysis
What the AI agent can set for logic, design and languages
An AI agent on the empirio.ai MCP server does more than line up questions. It also sets the rules that decide who sees which question at all, and it sets the look, the page layout and the languages. This is the part most readers underestimate.
For logic the agent works with the same rules as the editor. A rule checks the answer to one question and then triggers an action: jump to a question, show or hide a question or a whole page, or end the survey. The comparison is equals, not equals, contains, greater than or less than.
| Setting | What the AI agent can set |
|---|---|
| Logic | jumps, showing and hiding, ending the survey |
| Page layout | all on one page, one question per page or pages of your own |
| Design | one of eight templates, accent colour, logo |
| Languages | main language and others of the nine country versions |
| Start and end | your own titles, text and a button with a target |
| Run window | start and end time |
It is still worth checking yourself whether the logic and branching in your survey make sense. One note on order: logic, design and languages belong in the draft before the first answer arrives.
Note
When the agent creates a translation, an entry may still carry the text of the main language and be marked as open. That text is a placeholder, not a finished translation. Every language deserves one read before you publish.
Data protection: what the AI provider sees and what stays in the EU
All survey data at empirio.ai sits on servers in the EU. EU hosting is included on the free plan and a data processing agreement is part of the free plan as well. For an organisation in the UK the relevant provision is Article 28 UK GDPR, which requires a written contract with any processor, and the supervisory authority is the ICO.
The MCP server changes none of that, but it moves one boundary. Everything the AI agent retrieves through the MCP server lands in its chat history and is therefore processed by your AI provider as well, under that provider’s own rules. The raw data stays stored in your empirio.ai account, but the extracts the agent pulled also sit in the chat.
What to do about it in practice
Decide per survey whether the AI agent may see the answers. Creating and publishing does not need them, analysing does. You grant the rights when you connect and individual ones can be left switched off. Creating surveys without access to the answers is therefore a sensible default, particularly for a staff survey that a line manager will act on. The legal frame for surveys is explained in the guide to data protection in surveys.
The MCP server gives the AI agent the rights of your account. No more than that, but no less either.
Common mistakes when running surveys through an AI agent
The connection is quick, the mistakes come later. Four of them turn up often enough to deserve their own lines. All four share one cause: the AI agent answers so confidently that nobody checks the maths.
None of the four is a fault in the technology. The MCP server delivers exactly what it was asked for and the gap opens between the question in the chat and what ends up in the report. The same remedy works at every point: check a number before you pass it on.
Mistake 1: trusting the export without asking
An export through the empirio.ai MCP server contains completed participations only by default. Anyone analysing a survey with a lot of drop-outs gets a flattered file, with no note anywhere saying so. Tell the agent explicitly whether incomplete participations should be included, and compare the row count against the completion rate from the statistics.
Mistake 2: passing on percentages from a multiple-choice question
For a multiple-choice question the analysis counts selections, not people. Twelve people can tick 30 boxes and a percentage built on that does not add up to 100. Ask the AI agent for the number of participants in this case and write it into every report that may be quoted later.
Mistake 3: missing the placeholder translation
A newly created language version can still hold the text of the main language. In the chat it looks like a finished translation, in the survey English text appears under a French language option. Check every language once in the preview before the survey goes live.
Mistake 4: granting delete rights you do not need
An AI agent on the empirio.ai MCP server can also delete surveys and individual answers, provided you grant that right. A loosely worded instruction such as “clear out the test data” may then hit real answers. Leave the delete rights switched off while you do not need them and grant them for the one occasion when you really are tidying up.
Checklist before publishing
- The agent has checked the plan status of the survey.
- Every language version has been read once in the preview.
- The logic has been clicked through with a test participation.
- Delete rights are granted only where they are needed.
- It is settled whether drop-outs count in the analysis.
Conclusion
The empirio.ai MCP server pays off when a large part of your work already happens in an AI chat and you often need small surveys for a team, a club or your customers. For a single, carefully considered study you are faster and more precise in the empirio.ai editor, because you see every scale directly. For complex panel studies with quota control empirio.ai is not the right tool.
What comes next
- Want to automate whole workflows? Automate surveys with Zapier
- Need the survey in several languages? Set up a multilingual survey
- Still looking for the right tool? Survey tools compared
Want to try your AI agent straight away?
The free plan from empirio.ai is enough for a first survey through the MCP server. Register with empirio.ai for free