Skip to main content
Usertour is open-source user onboarding with an MCP server built in: connect an AI assistant like Claude Code, Cursor or Codex, describe the experience you want, and it authors, themes and publishes the content directly in your project — matched to your app’s design, verified in your running app. The video below does exactly that against a real app (shadcn-admin): four prompts, and after each one — the result live in the app, then the content and data in the Usertour dashboard.

Before you start

All you need is a Usertour account and an AI assistant connected to the MCP server. With Claude Code that’s the plugin (it registers the MCP connection, the authoring skills, and the SDK-install skill in one):
Then run /mcp and authorize in the browser — pick the project, the environments, and the access level. That’s the whole setup: you don’t even install the Usertour SDK by hand — on the first prompt the assistant notices it’s missing and wires it into the app itself, then builds the content.
Self-hosting? The plugin defaults to Cloud — point it at your own instance before connecting (your exact URL is shown in Settings → MCP): export USERTOUR_MCP_URL="https://<your-usertour-host>/mcp"
Optional, recommended: also install the Chrome DevTools plugin so the assistant can open your running app, pick element selectors from the real DOM, and verify each experience actually renders before publishing — noticeably better tooltip accuracy.
Using Cursor, Codex, VS Code or another client? See the MCP connection guide — every client gets the same one-URL setup, and Settings → MCP in the app shows the steps with your instance’s URL pre-filled.

Step 1 — A feature onboarding flow

In the app — before building anything, the assistant notices the app has no Usertour SDK yet and installs it on its own (the loader snippet, init() and identify()); then the flow runs against the real UI: tooltips anchor to the actual Create Task controls, and the styling matches the app because the agent read the design system first and themed the content to it. In the dashboard — the flow appears under Content with its steps, targeting and a published version; Analytics starts counting views and completions as sessions come in.

Step 2 — An adoption checklist

In the app — a checklist panel lists the User List features as tasks; completing a task can launch the matching flow, so the two content types work as one adoption journey. In the dashboard — the checklist shows per-task completion in Analytics, so you can see which features users actually adopt and where they stall.

Step 3 — A product feedback survey

In the app — a multi-question survey renders in the app’s own look: ratings, scales and open-text questions in a sensible order. In the dashboard — every question gets its own breakdown under Analytics: score distributions for the ratings, and the open-text answers collected per response, ready to read through (or to hand back to the AI to summarize).

Step 4 — A welcome announcement + resource center

In the app — new users get a speech-bubble welcome pointing them at the User List page, and the resource center becomes the lasting home for everything built above: the flow, the checklist and the survey, organized into groups and launchable on demand. In the dashboard — the announcement and resource center sit alongside the rest under Content, each with its own targeting and analytics, and Publish history records what went live in which environment.

One more: a one-prompt task

Not everything is a four-step journey — one-off tasks are a single prompt. Here the assistant builds a video modal, verifies the embed actually plays, and leaves it as a draft for review instead of publishing:

Where to next

Analyze your onboarding with AI

The other half of the loop — five prompts that turn the collected data into rankings, funnels and survey readouts.

Quickstart

Install the SDK and identify your users — everything above builds on it.

MCP server reference

Connect any client, see every tool, understand scopes and read-only mode.

Build in the visual builder

Prefer hands-on? Build the same flows, checklists and surveys visually.

Self-host Usertour

Run the whole platform — MCP server included — on your own infrastructure.