Getting started
What ScreenPilot does, how a video gets made in four steps, and your first video in three asks to your AI app.
ScreenPilot lets your AI app make a screen video for you. It records your screen, or uses your apps itself, then edits the footage into a finished MP4 with zooms, captions and an AI voiceover. You do not edit anything and you do not set up a screen recorder.
What ScreenPilot does#
ScreenPilot works inside the AI app you already use: Claude, ChatGPT Codex, Cursor, OpenCode, Gemini CLI, Copilot, Windsurf, VS Code and many more. It runs as an MCP server on your computer and gives the AI app three things: eyes on your screen, a recorder, and a video editor.
- No API key and no account. Your AI app does the thinking. ScreenPilot does the recording and editing.
- Everything runs on your computer. By default the voiceover is the one thing that needs the internet: it uses Edge voices in English, Arabic and many other languages. You can install free voices that run on your computer instead (see Voices).
- It records any app. Native apps, code editors, terminals and browsers all work, on Linux, Windows and macOS.
Not installed yet? Start with Install ScreenPilot, then come back here.
How it works#
Every video goes through the same four steps. Your AI app runs them for you.
- Record or drive. ScreenPilot records the screen along with pointer, click and active-window events. You can do the demo yourself while it records, or let the AI app do it: it looks at your screen, then moves the mouse, clicks and types on its own.
- Analyze. The AI app looks at a contact sheet of the recording and at single frames with a coordinate grid. It finds idle time, clicks and the moments that matter.
- Storyboard. The AI app writes a
storyboard.jsonfile: which parts to keep or speed up, where to zoom, which captions and highlights to show, and the voiceover lines. - Render. ScreenPilot builds the MP4 from the storyboard. It renders a quick draft first, checks frames of the result, fixes what looks wrong, then renders the final video.
Your first video in three asks#
Restart your AI app after installing, then type these in order. Change the wording to fit your app.
1. Check the setup#
Run the ScreenPilot doctor and tell me if anything is missing.This checks the recorder, permissions and voices, and says how to fix anything that is not ready. On a Mac or on Linux with Wayland, your system may ask for permission here. See Install on macOS and Install on Linux.
2. Make a video#
Make a demo video of our dashboard: open it, show the reports page, and add an English voiceover.The AI app opens the page, performs the clicks while recording, and writes the storyboard. Keep your hands off the mouse and keyboard while it works: touching the mouse stops the action. The first video takes longer because the video renderer warms up.
Prefer to show it yourself? Ask this instead:
Record my screen while I show the signup flow, then make a short video with captions.Do your demo, then tell the AI app you are done.
3. Change it#
Make it shorter, cut the waiting parts, and add an Arabic voiceover.The AI app edits the storyboard and renders again. Ask for as many changes as you like: a different voice, bigger zooms, a title card, background music, a vertical version for Reels or TikTok.
You can also start from a video you already have:
Turn this recording into a tutorial with zooms on every click: ~/Videos/demo.mp4Where your videos are saved#
Recordings and finished videos go in the ScreenPilot folder inside your Videos folder. On a Mac, that is the Movies folder. The AI app tells you the exact path when a video is ready.
- Windows: the ScreenPilot folder inside your Videos folder (the one Windows uses, even if OneDrive moved it).
- macOS:
~/Movies/ScreenPilot - Linux: the ScreenPilot folder inside your Videos folder, which is
~/Videos/ScreenPilotunless your desktop uses another name for Videos.
To save somewhere else, run the installer's setup with --output <dir>. See the command line guide.