I Let Claude Code Write a 20-Lesson Tutorial Course for My App From Its Codebase

I build dashboards. Admin panels, reports, forms, the usual things people need at work. Writing them is the part I enjoy. Explaining them is the part I don't.

Every project ended the same way: "how do I add a product?", "where is the monthly export?", the same screen share again for each new person on the client's side. I knew I should make tutorial videos, but recording them meant a quiet room, a clean screen, retakes every time I misclicked, and then the next release would make half of them wrong.

So I tried something else. I opened Claude Code in my project folder and gave it one prompt asking it to write a tutorial series for the app. It already had the whole codebase in front of it: every page, every button label, every role, every form field. It came back with a series of more than 20 lessons. I imported that reply once into DashCast, the tool I've been building for this, and each lesson became a narrated video with clicks, zooms and subtitles. I didn't record anything.

That changed how I think about tutorial videos, so here is what I learned, and what the tool does and doesn't do.

The AI in your project already knows your app

The coding assistant you already use has read your routes, templates and labels. That makes it a good author for "how to use this" lessons, better than anything looking at your app from the outside.

The flow is: DashCast gives you a prompt, you paste it into Claude Code, Cursor or any AI tool opened in your project folder, and you paste its reply back. A series is imported in one step. Rendering still happens lesson by lesson.

Your source code doesn't leave your machine. The prompt runs in your own tool, and DashCast only receives the finished lessons.

No repo? Start from a link

Sometimes the app isn't yours: a client's system, a third-party tool, an old project you no longer have the code for. In that case you give DashCast a link, a test login and one sentence about the video you want. A built-in AI writer explores the screens, writes the lesson and checks each click before you render. It is limited to 3 lessons a day on a free account and 10 on paid plans.

A script, not a recording

This is the part that mattered most to me. A recorded video is frozen on the day you record it. A lesson here is a script: what the voice says and what happens on screen at each moment.

When a button moves or a menu gets renamed, you change that one step and render again. The voice, the timing and the rest of the lesson stay the same, and only the lessons that changed need a new render.

If you want finer control, there's a timeline view (the Scene Designer) that shows the narration word by word with the actions on their own track. You drag a click to the word where it should happen.

Checking which lessons a release broke

After a deploy, there's a "Check selectors" button on each lesson. In about a minute it lists every button and field the lesson can no longer find on the live app. I use it as a quick way to see which parts of the UI a release actually changed, before a user watches an outdated video.

Other things I use it for

• Client handover. Instead of a call after delivery, I hand over a short course with the code: "How to add a product", "How to approve an order", "How to export the monthly report". It's also something you can list in a quote.
• Release videos. People skip changelogs but will watch 60 seconds of a new feature. There's a "What's new" video type for that.
• Support answers. One common ticket, one video under a minute, sent as a link instead of a long reply.
• A launch video for your own product. The "Marketing" type makes a 30 to 60-second video that talks about outcomes rather than buttons.

Logins and sensitive data

Most of what users need to learn is behind a login. Lessons sign in with a test account before recording starts, so the password never shows up in the video or the narration. Fields like emails, phone numbers or amounts can be blurred. I use a test account with sample data so the videos show what a real user sees without exposing anything real.

Languages

The same lesson can be rendered in 14 languages: English, Arabic, French, German, Spanish, Italian, Portuguese, Dutch, Russian, Turkish, Indonesian, Persian, Urdu and Hindi. The clicks stay the same and only the narration changes. Arabic, Persian and Urdu get right-to-left subtitles. Each video downloads as an MP4 with its subtitle file. Chinese, Japanese and Korean narration aren't available yet.

Teams

On the Pro and Business plans a team shares one workspace with admin, editor and viewer roles. Pro has 2 seats and Business has 5.

What it costs

You pay for rendered minutes. Voice, subtitles and languages are included, writing lessons with your own AI tool costs nothing extra, and failed renders aren't charged.

Where it doesn't fit

• It records web apps in a browser. It doesn't do desktop or mobile apps.
• It isn't for talking-head videos of yourself.
• If your UI changes completely, you'll still be rewriting lessons. Small changes are cheap; redesigns aren't free.

FAQ

Do I need video editing skills?
No. You describe the video, the AI writes the lesson, and you get a finished MP4 with voice, subtitles, clicks and zooms. If you want to adjust timing, the Scene Designer works like a simple timeline.

Does DashCast see my source code?
No. When you use your own AI tool, the prompt runs on your machine inside your project folder, and DashCast receives only the finished lessons. You can also skip the code entirely and start from a link to your app.

Does it work with my stack?
It works with any web app that opens in a browser, whatever it's built with: Laravel, Django, Rails, Node, React, Vue, WordPress and so on. It doesn't record desktop or mobile apps.

Can I make videos for a client's app?
Yes. Use a test account on the client's app, render the lessons, then download the MP4 files and subtitles or publish them on a page for the client's users.

This post was first published on the DashCast blog: DashCast for Developers.