MCP server
MCP is the standard AI assistants use to operate other programs. Analyze responses in your assistant, create surveys from there, and grant only what the job needs.
Workflows you can build with it
Each example is a workflow you can set up with an automation service, an AI assistant or your own application.
Analyze results in your assistant
Once the account is authorized, the assistant answers with averages, completion rates and participant counts for a running survey. For teams using neither ChatGPT nor Claude but the assistant their company settled on.
Sort free text into themes
Written answers reach the assistant in full, and it builds themes with counts and examples out of them. For studies with hundreds of comments nobody reads to the end.
Create a survey from the assistant
A description of the goal produces a complete draft in empirio.ai, with questions and options the assistant keeps editing on request. For departments preparing a study in the tool they already work in.
Grant only what the job needs
Approving surveys but not responses leaves a connection that can never read an answer. For accounts where more is at stake than one study.
Build the assistant yourself
Your own agent speaks to the same address a packaged client does. For engineering teams wiring a language model into an existing workflow.
What the integration transfers
Each point is a separate example in the direction of the arrow. Configure each workflow independently and link the relevant survey for actions on an existing survey.
empirio.ai → MCP server
- Surveys and results for the account that authorized
- Exports as a download link that expires after an hour
- Written answers in full, ready to be summarized
MCP server → empirio.ai
- Changes to a survey, once the account allows it
- Creating, editing and publishing from any client
- A survey published or closed, on a word in the assistant
How to set it up
Copy the MCP server URL from your integrations settings, add it in the assistant, and confirm the grant once in the browser. That grant decides what the assistant reaches: the surveys, the responses or both. Delivery addresses are deliberately out of reach here — the REST API manages those.