Reporting

The fastest way to generate reports from Confetti is through the MCP server — connect Claude (or another MCP client) and ask questions in plain language. No API keys, no scripts, no CSV exports.

Get started

1. Connect the Confetti MCP server in your AI client.
2. Ask a question about your surveys or responses — the assistant calls the right tools for you.

You only see Teams and Surveys you already have access to in Confetti. The MCP server is read-only — it never creates, edits, or deletes data.

What you can ask

The MCP server exposes tools for listing surveys, reading individual responses, and fetching score distributions. Your assistant combines them based on your question.

Score breakdowns (CSAT, ratings)

Ask about rating questions over a time window — the same data as the Insights tab score breakdown.

Example prompts:

• "What are the score distributions for my Product Feedback survey over the last 30 days?"
• "Compare this quarter's CSAT to last quarter for the onboarding survey."
• "Show me the rating breakdown for question X in the last 7 days."

The assistant uses get_score_distributions with a timeRange such as last-7-days, last-30-days, q1-2026, or custom_2026-01-01_2026-01-31. See Tools for the full list.

Response summaries

Ask the assistant to read and synthesize recent responses.

Example prompts:

• "Summarize the last 20 responses to the CSAT survey."
• "What are people saying about page load times in recent feedback?"
• "List responses from the last week where the rating was 1 or 2."

The assistant uses list_responses (paginating as needed) and get_survey to resolve question titles.

Survey overview

Example prompts:

• "Which of my surveys have the most responses?"
• "List my active surveys and their response counts."
• "Show me the questions on the Product Feedback survey."

The assistant uses list_teams → list_surveys → get_survey.

Example: weekly CSAT digest

Instead of writing a script, ask your assistant:

"For my Product Feedback survey, give me a weekly CSAT summary: total responses, score distribution for the last 7 days, and a brief summary of written feedback from low ratings."

The assistant will:

1. Find the survey via list_surveys.
2. Get question IDs from get_survey.
3. Pull score distributions with get_score_distributions (timeRange: last-7-days).
4. Fetch recent low-rated responses with list_responses and summarize them.

You can paste the result into Slack, email, or a doc — or ask the assistant to format it for you.

MCP vs other approaches

What MCP does not cover yet

Engagement funnel metrics — views, dismissals, unique viewers — are only available via the GET …/stats API endpoint or the dashboard Insights tab. If you need those numbers in an automated pipeline, use the API reporting guide.

For everything else — response data, score distributions, and natural-language summaries — start with MCP.

Tool reference

See Tools for inputs, outputs, and the typical list_teams → list_surveys → get_survey flow.

Not connected yet? See Connecting.