Record your screen, point by drawing and speak, and hand it off to AI agent. Video prompts for Cursor, Claude, and Codex or any AI coding agent — local-only and free.
The short version: it’s a Free, local-only macOS app that turns screen recordings into agent-ready prompts for Cursor, Claude, Codex, or any AI agent. Think of it as sharing your screen and annotating your thoughts to talk a developer through what to build.
I wrote long, detailed prompts to get better results, but the agent still guessed wrong. Screenshots didn't fix it either—they’re too static to capture movement, form input flows, or real-time UI feedback.
So I built the exact solution I wanted, I use it on my daily tasks now, and it’s been a game-changer. It saves me so much time—I explain a task once, and it’s done.
How it works:
1. Start Recording: Press Ctrl + Shift + R or click record button.
2. Annotate Your Screen: Press Ctrl + 1 or 2 ... to toggle drawing tools. Standard drawings clears automatically,
but holding Shift while drawing keeps them permanently on screen—perfect for sketching layouts or UI flows.
3. Stop Recording: Press Ctrl + Shift + R again or click the Stop button.
4. Open in Cursor, or copy the session prompt into your preferred AI agent and let it cook.
No cloud upload. No login. Recordings stays on your Mac.
Features
Consumes fewer tokens: The agent reads recordings via Local MCP using keyframes and speech—not a giant video dump.
Multi-monitor support: Point or draw across different screens, switch context seamlessly, and let the AI understand what you mean.
Secure - Recordings stays in your device, we don't upload your recordings. Speech transcripts and frames are stored in your device locally.
Agent-agnostic: Works with any AI coding assistant (Cursor, Claude, Codex, and more).
Annotation tools: Easy-to-use drawing and highlighting tools so you can explain your idea clearly to the AI.
If you already vibe with AI coding agents and hate writing long prompts, I’d love for you to try it and tell me where it breaks.
Happy to answer anything — let me know how you can use it on your daily work flow,
@emrreguney That’s great to hear. Thank you. Happy to build something useful! 😁
Report
Congrats on the launch! Describing a UI problem in text is usually the slowest step when I work with a coding agent, so sending keyframes and a transcript rather than the video is the part that makes this practical. On a multi step flow, does the transcript stay tied to the frames in order?
Thanks@alieksia, Yes, the transcript is synchronized frame-by-frame along the timeline. Every word is tied to a specific timestamp, so if you say "change this button position" while drawing a line, that exact moment aligns with the corresponding frame.
Report
Honestly, being able to sketch a feature on screen while explaining it sounds so much better than writing text! That would really feel like showing an idea to a colleague sitting right next to you. Congrats on the launch! :)
@veronica_macadam Exactly! Prompts are often ephemeral—once you get the output, you move on. Keeping everything local gives you that speed while protecting privacy and security.
Report
Love the local-first approach! Are there plans to support Windows or Linux in the future?
@busmark_w_nika Yes, you're right. First we typed prompts, then we used speech-to-text (voice as prompts), and now we have video as prompts. I think we're moving toward a much more natural way of communicating to AI—just like how we talk to humans. An AI that can also respond with a video showing what to do would be a great idea. 🤔
Annotate
Hey Product Hunt 👋
I’m the maker of Annotate.
The short version: it’s a Free, local-only macOS app that turns screen recordings into agent-ready prompts for Cursor, Claude, Codex, or any AI agent. Think of it as sharing your screen and annotating your thoughts to talk a developer through what to build.
I wrote long, detailed prompts to get better results, but the agent still guessed wrong. Screenshots didn't fix it either—they’re too static to capture movement, form input flows, or real-time UI feedback.
So I built the exact solution I wanted, I use it on my daily tasks now, and it’s been a game-changer. It saves me so much time—I explain a task once, and it’s done.
How it works:
1. Start Recording: Press Ctrl + Shift + R or click record button.
2. Annotate Your Screen: Press Ctrl + 1 or 2 ... to toggle drawing tools. Standard drawings clears automatically,
but holding Shift while drawing keeps them permanently on screen—perfect for sketching layouts or UI flows.
3. Stop Recording: Press Ctrl + Shift + R again or click the Stop button.
4. Open in Cursor, or copy the session prompt into your preferred AI agent and let it cook.
No cloud upload. No login. Recordings stays on your Mac.
Features
Consumes fewer tokens: The agent reads recordings via Local MCP using keyframes and speech—not a giant video dump.
Multi-monitor support: Point or draw across different screens, switch context seamlessly, and let the AI understand what you mean.
Secure - Recordings stays in your device, we don't upload your recordings. Speech transcripts and frames are stored in your device locally.
Agent-agnostic: Works with any AI coding assistant (Cursor, Claude, Codex, and more).
Annotation tools: Easy-to-use drawing and highlighting tools so you can explain your idea clearly to the AI.
If you already vibe with AI coding agents and hate writing long prompts, I’d love for you to try it and tell me where it breaks.
Happy to answer anything — let me know how you can use it on your daily work flow,
Thanks for being here.
Boost CTOR
@kim_ben_g Great idea, i can feel it will be a daily product for me. thanks for building and sharing!
Annotate
Congrats on the launch! Describing a UI problem in text is usually the slowest step when I work with a coding agent, so sending keyframes and a transcript rather than the video is the part that makes this practical. On a multi step flow, does the transcript stay tied to the frames in order?
Annotate
Thanks@alieksia, Yes, the transcript is synchronized frame-by-frame along the timeline. Every word is tied to a specific timestamp, so if you say "change this button position" while drawing a line, that exact moment aligns with the corresponding frame.
Annotate
Thanks @etiennegarcia 🙌 , Exactly! It feels just like sharing your screen and marking things up while explaining an idea to a colleague.
Love that it stays completely local — no upload step means the friction between recording an idea and getting a usable prompt basically disappears.
Annotate
@veronica_macadam Exactly! Prompts are often ephemeral—once you get the output, you move on. Keeping everything local gives you that speed while protecting privacy and security.
Love the local-first approach! Are there plans to support Windows or Linux in the future?
Annotate
Hi @kim_ben_g are there any plans for Android & Windows too ? Love to hear that.
Annotate
minimalist phone: reduce your screentime
I think that this is the future and also outputs (answers by AI) that will show you on the video what to do.
Annotate
@busmark_w_nika Yes, you're right. First we typed prompts, then we used speech-to-text (voice as prompts), and now we have video as prompts. I think we're moving toward a much more natural way of communicating to AI—just like how we talk to humans. An AI that can also respond with a video showing what to do would be a great idea. 🤔
minimalist phone: reduce your screentime
@kim_ben_g TBH, I would need a video showing, as I do not understand some text-based steps.
Annotate
@busmark_w_nika Absolutely, a screenshot is worth a thousand words but a screen video is worth more, I would say. 😂