Ask your assistant what users said

Survey results usually live in a dashboard you have to remember to open. Then you export them, paste them somewhere, and try to summarise two hundred open answers by hand. We wanted the results one question away instead: in the assistant you already have open while you work.

So Respondo runs a Model Context Protocol server. MCP is an open standard for connecting assistants to tools and data. Connect Claude, or any client that speaks it, and your assistant can read what people answered and shape what they are asked next — in the same conversation where you are writing the roadmap or the release notes.

Connect in one line

The server lives at one URL. In Claude Code, add it with a single command:

claude mcp add --transport http respondo https://app.myrespondo.com/api/mcp

Other clients add it as a remote HTTP server using the same URL. The first time your assistant calls a tool, a Respondo consent screen opens in your browser. Approve it and you are connected — there is no key to copy and nothing to paste into a config file.

Eight tools

Once connected, your assistant has eight tools. Four read what people said:

• get_results — totals for one survey: response and completion counts, and per question the answer split, the average for a rating, the NPS score for an NPS question, and up to 25 sample open answers. This is the one to reach for when the question is “what did people say?”
• list_responses — individual answers, newest first, for when one respondent's reply matters. It can include the email and name if your site passed them.
• list_surveys — every survey with its status, question count and response count.
• get_survey — one survey in full: its questions and its display, targeting and appearance settings.

And four shape what people are asked:

• create_survey — drafts a new survey. It is always created as a draft, never live.
• update_survey — changes a survey's name, questions or settings.
• set_survey_status — publishes, pauses, closes or returns a survey to draft.
• delete_survey — removes a survey and every response it collected.

From a question to a follow-up survey

Here is how a typical exchange goes. You do not name tools; you ask in plain language and the assistant picks them.

1. 1You ask: “What did people say about onboarding this month?” The assistant calls get_results on your onboarding survey and summarises the split and the common threads in the open answers.
2. 2You ask: “Draft a follow-up for people who found it difficult.” The assistant calls create_survey. The new survey is a draft, so nobody sees it yet.
3. 3You read it back, ask for a change or two, and the assistant calls update_survey.
4. 4When you are happy, you say so, and the assistant calls set_survey_status to publish it.

A few more questions that work well:

• “Summarise the open answers on the cancellation survey.”
• “Which survey has the lowest completion rate?”
• “Draft an NPS survey for the billing page, shown to 10% of visitors.”
• “Pause every survey that is live.”

You stay in charge

Giving an assistant access to your product data should not mean giving it the run of the place. The server is built around that:

• Signs in as you. The consent screen is Respondo's own, and the connection sees what your account sees — never another account's surveys or responses.
• Two scopes. Access is granted as surveys:read and surveys:write, so a connection can be allowed to read without being allowed to change anything.
• Confirm first. The tools that publish and delete tell your assistant to check with you before it calls them, and new surveys always start as drafts.
• Revocable. Every connected client is listed in Settings under Connected apps, and any of them can be revoked there.

Sign-in uses OAuth with PKCE, and every token is issued for the Respondo server alone. The full details are on our security page.