Recording by hand
Open the screen recorder, rehearse the flow, click too fast, click the wrong thing, start over. Then do it again for mobile.
Klevo · demo videos of your web app, recorded by the AI you already use
Your app, demoed by AI.
Klevo is an MCP server. Connect it once and "record a demo of the signup flow" comes back as an MP4: your AI clicks through the app, Klevo records it and renders a smooth cursor, zooms on the key clicks, captions and a device frame. No screen recorder, no video editor.
01The problem
Everyone needs a demo video. Nobody wants to make one.
The landing page, the Product Hunt post, the changelog, the onboarding email, the support doc: each wants a short clip of the app doing the thing. Each clip costs an afternoon, and the UI changes next week.
Editing in a video tool
The raw recording is flat: no cursor you can follow, no zoom on the button that matters, no captions. Making it watchable means an editor and an hour.
Demos that go stale
The clip on the landing page shows last quarter's sidebar. Re-recording is the real cost, so the stale one stays up.
02How it works
You describe the flow. The video comes back.
Klevo connects through MCP, the standard Claude, ChatGPT and Cursor already support. Your AI drives your app one step at a time, looking at a screenshot after each click, and Klevo turns the steps into a video.
01
Connect once
Add Klevo as a connector in Claude or ChatGPT and sign in. Cursor and Claude Code take a key. If your app has a login, save a test account as a secret: Klevo logs in before recording starts.
02
Describe the flow
"Record a demo of creating a project on staging.myapp.com, desktop, 16:9." The AI opens the page, clicks and types through the flow, writes a caption for each step and finishes.
03
Get the MP4
Klevo replays the steps in a clean browser and renders the cursor, zooms, captions, device frame and intro card. About a minute later the download link is in the chat. Need a vertical cut? "Render it 9:16 on a phone." No re-recording.
03What you get
A finished video, not a screen recording.
The exploratory session your AI runs is never the video. Klevo replays the saved steps in a fresh browser and composes the clip from scratch, so there are no pauses, no mis-clicks and no cursor hunting for the button.
Cursor, zoom, transitions
A smooth cursor on desktop, a touch ripple on mobile, an automatic zoom into the click that matters, and on Pro three styles (smooth, snappy, minimal) and an optional click sound.
Devices and formats
Record as desktop, tablet or mobile. Export 16:9 for the landing page, 9:16 for stories and shorts, 1:1 or 4:5 for the feed. The two choices are independent.
Device frames
A browser window on desktop and an iPhone on mobile, on a background in your brand color, with an intro and an outro card. Pro picks any frame or none, any Google Font, its own background, and hides the cards.
Captions
Your AI writes one line per step; the video shows it, timed to the step, in the language the AI wrote it. A natural voice of your choice can read them aloud on voice credits, and on Pro you can edit or hide them.
Login without the login
Save a test account, a browser session or a magic link as a secret. Klevo signs in before recording, and password fields are blurred in every frame.
A failed step is a bug report
If a button is gone, the render stops at that step and hands back the step number, a screenshot and the page's console errors.
The demo is a script. Scripts don't go stale.
Every video comes from a small JSON file: the URL, the device, a reference to the login secret and the steps with their captions. Keep it in your repo. When the UI changes, tell your AI to render it again and a fresh browser repeats the same steps on the live app. One script, every format, forever current.
Secrets are referenced by name and never written into the script. Targets are roles and labels, not pixel positions, so a moved button still gets clicked.
{
"version": 1,
"url": "https://staging.myapp.com/projects",
"device": "desktop",
"auth": { "type": "password", "secret": "DEMO_LOGIN" },
"output": { "format": "16:9", "frame": "browser", "style": "smooth" },
"steps": [
{ "action": "click", "target": "role=button[name='New project']",
"caption": "Start a new project", "zoom": true },
{ "action": "type", "target": "label=Name", "text": "Q4 launch",
"caption": "Give it a name" },
{ "action": "click", "target": "text=Save", "caption": "Done" }
]
}
04Works with
The AI you already pay for.
Nothing to install. Klevo is a remote connector: it works in Claude on the web, on the desktop, on your phone and in ChatGPT, and your AI, on your plan, does the driving.
Claude
Claude.ai, the desktop app, the mobile app and Claude Code. Add a custom connector and sign in.
ChatGPT
A custom connector in ChatGPT's settings. Sign in with your Klevo account.
Cursor and VS Code
One block in your MCP settings file, with a key from your account.
Anything with MCP
Other AIs that support remote MCP servers connect the same way.
05Pricing
Start free. Go Pro when the demos ship.
Free renders 3 videos a month at 720p with a small klevo.si mark in the corner. Pro takes the mark off, renders 1080p, unlocks the look (frame, fonts, background) and the step editor, and takes demos up to 120 steps. Ultra takes full product tours up to 240 steps, with the most voice credits. Narration runs on voice credits: Free starts with some, and packs never expire.
$0
Try it on the app you already have.
- 3 videos a month
- 10 voice credits to start, for narration
- Up to 20 steps per video
- 720p, every format
- Captions, intro and outro cards in the default look
- Small klevo.si watermark
$149/ year
Demos that never go stale. Yearly saves $79.
- 1,000 videos a month
- 200 voice credits a month, unused ones roll over up to 600
- 1080p, no watermark
- Device frame on or off, any Google Font, your own background
- Narration with your choice of voice and tuning
- {"Edit the steps" => "actions, targets, captions, waits and how long each shows"}
- {"Caption style" => "position, size, colors, no box, shadow"}
- Click sounds, hide captions, intro or outro
- Up to 120 steps per video
- Re-render any script in another format or device
$399/ year
Full product tours, narrated. Yearly saves $189.
- Everything in Pro
- Up to 240 steps per video, for full product tours
- 500 voice credits a month, unused ones roll over up to 1,500
- Recordings stay open up to 2 hours
- 50 saved logins, 5 recordings open at once
Voice credits
Narration, paid by the character.
One credit reads 1,000 characters with the Fast voice (Advanced spends 2). A typical demo spends about 1. Free starts with 10 once. Pro gets 200 a month and Ultra 500, rolling over up to 3 months' worth while you subscribe. Re-rendering the same captions is free. Packs never expire.
06FAQ
What is an MCP server?
MCP (Model Context Protocol) is how AI assistants use outside tools. Once Klevo is connected, your assistant calls it by itself when you ask for a demo, the same way it searches the web or reads a file. You never leave the chat, and there is nothing to install.
What do I get back?
An MP4 with a smooth cursor (or a touch ripple on mobile), an automatic zoom into the key clicks, a caption per step, a device frame and an intro card, plus a download link in the chat. The video and its script are kept in your account. Free renders at 720p, Pro at 1080p.
How does Klevo log in to my app?
You save a secret once: a test account's login (email and password), a browser session exported from Playwright, or a magic link. Klevo signs in with it before recording starts, so the login never shows in the video. Use a test account, not your own.
Is my password safe?
Secrets are encrypted at rest, never shown again after saving, and sent only to the recording service for a render or a live session. They are referenced by name in scripts and never written into them. Password fields are blurred in every frame, and the usage log keeps request sizes, never what was typed.
What about two-factor, Google sign-in or a captcha?
They can't be automated, by design. The way around all three is the same: log in once in a browser, export the session (Playwright's storageState) and save it as a secret. On staging, a test-login link saved as a magic link is the most durable option.
Can I record an app on localhost?
No. Klevo's browser runs on our servers and only opens public URLs, so it needs a staging or production address it can reach. Private and local hosts are refused.
How long does a render take?
About a minute for a short demo. The AI's exploration happens live while you watch the screenshots; the video itself renders in the background, and your AI (or the Videos page) tells you when it's ready.
Can I get the same demo in another format?
Yes, without recording again. Every video comes from a saved script; ask your AI to render it 9:16 for a story, 1:1 for the feed, on a phone frame or in another style. Each render counts as a video for the month.
What if a step fails?
The render stops at that step and reports which one, with a screenshot of the page and its console errors. Either the demo needs a different target, or your app has a bug: both are useful to know. Failed renders don't count against your month.
What is the watermark on Free?
A small klevo.si mark in the bottom-right corner of the video. Pro renders without it; after upgrading, render the script again and the clean version comes back.
What does Pro add?
1000 videos a month instead of 3, 1080p without the watermark, demos up to 120 steps instead of 20, and more saved logins. Pro also unlocks the look and sound: any device frame or none, the three styles, any Google Font, your own background, click sounds, editing the steps, and hiding the captions, intro or outro, plus 200 voice credits a month. Free renders every format in the default look.
What does Ultra add?
Everything in Pro, for full product tours: demos up to 240 steps, recordings that stay open up to 2 hours, 50 saved logins, and 500 voice credits a month that roll over up to 1500. On Pro already? Upgrade from your account: Pro ends when Ultra starts, and what's left of it is credited to your next bill.
How do voice credits work?
Narration reads your captions with a natural voice and spends voice credits: one credit reads 1,000 characters with the Fast voice, two with Advanced, and a typical demo spends about one. Free starts with 10, once. Pro gets 200 a month that roll over up to 600, Ultra 500 up to 1500; monthly credits last while you subscribe. Rendering the same captions again in another format is free. Need more? Packs of 50 for $4 or 200 for $12 never expire.