Quickstart

Three steps. Provision a project, drop the component into your app, ship a survey.

For agents — single self-contained blob

Drops into Claude / Cursor / ChatGPT and reproduces the full install end-to-end.

1. Provision a project

From an agent, call the MCP tool provisionProject. From a human, sign in at /sign-in and create a project from the dashboard.

// MCP — POST /api/mcp
{
  "jsonrpc": "2.0",
  "id": 1,
  "method": "tools/call",
  "params": {
    "name": "provisionProject",
    "arguments": { "name": "my-app" }
  }
}

// → {
//   "projectId": "proj_...",
//   "publishableKey": "pk_...",
//   "serviceKey": "sk_...",
//   "dashboardUrl": "https://getsurvey.dev/dashboard/proj_...",
//   "integrationSnippet": "..."
// }

2. Drop the component in

// app/feedback/page.tsx
import { Survey } from "survey";

export default function FeedbackPage() {
  return (
    <Survey
      projectId="proj_..."
      surveyId="srv_..."
      publishableKey="pk_..."
    />
  );
}

3. Read insights back

When responses come in, read summarized analytics with the summarizeResponses MCP tool — designed to drop straight into a downstream prompt.

// MCP
{
  "method": "tools/call",
  "params": {
    "name": "summarizeResponses",
    "arguments": { "projectId": "proj_...", "surveyId": "srv_..." }
  }
}

What you don't have to do

Create database tables for responses.
Build a survey editor or preview pane.
Wire up validation, submission, idempotency, retries.
Compute Likert means, NPS scores, free-text clustering.