Cursor that glides
A visible pointer eases to each button and types at a human pace — no instant jump-cuts, so viewers can follow the click.
Free · MIT · local · v0.2
Your AI films your product.
One prompt to a share-ready demo.
AIOProductOS Studio is a free, open-source MCP server that turns your AI host into your product videographer. Ask it to walk through your own web app — a cursor that glides to each click, camera zoom-ins, highlight callouts, storyline captions, an end card — and it renders a share-ready MP4 and GIF plus high-DPI screenshots. The videos PMs hate making become one prompt.
What it does
You type the story. It shoots the film.
Not a screen grabber — a director's kit. Studio drives a real browser through your product and frames every beat like a produced explainer.
Camera zoom-ins
Punch in on the detail that matters, hold, then pull back — the framing a real screen-recording never gives you.
Highlight callouts
Drop an accent ring and a label on the exact element you're talking about, so the eye lands where the story is.
Storyline captions
Narrate the walkthrough in on-screen captions — the explainer voiceover, minus the microphone.
Designed transitions
Scenes open behind a title card and change with a real transition, not a hard cut — it reads as a produced film.
End card + screenshots
Close on a branded end card, and pull high-DPI screenshots from the same run for your changelog and launch posts.
-
studio_startOpen a themed recording stage. -
studio_gotoNavigate behind a title card. -
studio_clickGlide the cursor and click. -
studio_typeType into a field, human-paced. -
studio_scrollEased scroll to reveal content. -
studio_captionShow a storyline caption. -
studio_narrateSpoken line — narration, no on-screen text. -
studio_highlightAccent ring + label on an element. -
studio_zoomCamera punch-in on a detail. -
studio_holdHold the frame for a beat. -
studio_end_cardClose on a branded end card. -
studio_screenshotCapture a high-DPI still. -
studio_cancelDiscard the take; keep screenshots. -
studio_finishStrip-black → voice/subs/crops → MP4.
New in v0.2 · Sound & social
Now it speaks — and posts everywhere.
The same run now carries a voice, captions you can burn in, a music bed, and every aspect ratio a social feed asks for. All set when you call studio_finish — and all still on your own machine.
Voice-over
100% local, offline — nothing uploaded.
Studio narrates your captions and studio_narrate lines with a
fully on-device voice (say → piper → espeak-ng).
No cloud, no API key, nothing leaves your machine. Pick a voice or use the OS default — timing auto-corrects
for the dark frames the trimmer removes.
Subtitles
.srt or burned-in.
Write a sidecar .srt from your captions, or render them into
the picture for muted-autoplay social. Cues never overlap.
Music bed
Looped, and ducked.
Lay a looped local audio track under the walkthrough — automatically ducked beneath the voice-over so narration always stays clear.
Social crops
One take, every aspect ratio.
Export 9:16 and 1:1 cuts — padded with your theme background — beside the 16:9 master. Reels, TikTok, Shorts, and square feed posts, all on-brand and in sync from the same run.
Why it's different
Cinematic, and yours — nothing leaves your machine.
Most "screen recording" tools are desktop grabbers you still have to edit. Studio is a director your AI host drives — and it runs entirely on your own machine.
Free
No account, no key
Install with npx and film. There's no sign-up, no paid tier, and no API key to paste.
MIT
Open source
Read the code, fork it, ship it in your own pipeline. MIT-licensed on GitHub.
Local
100% on your machine
Your browser, your URLs, your logged-in session via Playwright storageState — and even the v0.2 voice-over runs offline, on-device. Nothing is uploaded anywhere.
Any host
Works with your AI
Drive it from Claude, Cursor, Codex, or any MCP host — the host you already use.
Install
One line, and your AI can film.
Studio is published on npm. Run it directly with npx, or register it in your MCP host's config.
Run it
npx -y @aioproductoscom/mcp-studioClaude Code
claude mcp add studio -- npx -y @aioproductoscom/mcp-studioCursor / Codex — .mcp.json
{
"mcpServers": {
"studio": {
"command": "npx",
"args": ["-y", "@aioproductoscom/mcp-studio"]
}
}
}Requirements
Node 18+ · Playwright Chromium (npx playwright install chromium, one-time)
· ffmpeg on your PATH for MP4/GIF (without it you still get raw WebM).
Output is written to ./studio-output.
FAQ
Studio, answered
What is AIOProductOS Studio?
AIOProductOS Studio is a free, open-source MCP server that turns your AI host into your product videographer. You tell it — in plain language — to walk through your own web app, and it records a scripted film: a visible cursor that glides to each click, camera zoom-ins, highlight callouts, storyline captions, designed scene transitions, and an end card. You get a share-ready MP4 and GIF plus high-DPI screenshots. The demo clips, explainers, changelog GIFs, and launch screenshots that PMs hate making become one prompt.
Is it free?
Yes — completely free and open-source under the MIT license. There's no account, no API key, and no paid tier. Install it with npx and start filming.
Does it upload my screens anywhere?
No. Studio runs 100% locally — your browser, your URLs, your logged-in session (via Playwright storageState). Nothing is uploaded anywhere and there's no account to sign into. The rendered MP4, GIF, and screenshots are written to ./studio-output on your own machine.
Which AI hosts work?
Any MCP host. Studio speaks the Model Context Protocol, so you drive it from the host you already use — Claude, Cursor, Codex, and other MCP clients — by adding it to that host's MCP config. The host is where you type the story; Studio shoots the film.
What can it film, and what do I get out?
Anything you can reach in a browser and are logged into: a feature demo, a how-to explainer, a changelog GIF, a launch screenshot set. Studio's 14 tools cover a themed stage, cursor glide + type, eased scroll, captions, spoken narration, highlights, camera zoom, holds, an end card, and screenshots. studio_finish strips the black frames and renders a share-ready MP4 plus a GIF; you also get high-DPI PNG screenshots from the same run. With ffmpeg on your PATH you get MP4/GIF; without it you still get the raw WebM.
Can it add voice-over and subtitles?
Yes, both — set on studio_finish. The voice-over narrates your captions and studio_narrate lines with a 100% local, offline voice (macOS say, then piper, then espeak-ng). No cloud, no API key, nothing uploaded — it's the same privacy story as the rest of Studio. Subtitles write a sidecar .srt from your captions, or burn them into the picture for muted-autoplay social. You can also lay a looped local music bed under the voice, ducked automatically.
Can it export vertical or square video for Reels and TikTok?
Yes. Alongside the 16:9 master, studio_finish can export 9:16 and 1:1 cuts — padded with your theme background — for Reels, TikTok, Shorts, and square feed posts. It's the same run, so every format is on-brand and in sync.
Let your AI make the demo.
Free, MIT, and 100% local. Grab it on GitHub or npm and film your product in one prompt.