Install MCP
One npm command. The same protocol Cursor and Claude already speak.
Model Context ProtocolOne studio.
Every AI tool.
Connect Screen Beaver to Cursor, Claude, and any MCP-compatible client. Agents film product tutorials without you touching the screen — then polish and export locally.
Full MCP documentation →screenbeaver-mcp
$ npm i -g @screenbeaver/mcp
# then wire Cursor / ClaudeWorks with
>The difference
[1/7]
How it works
[2/7]Connect. Prompt. Ship the MP4.
Three steps — no editor timeline required until you want to refine.
Ask for a demo
The agent clicks, types, and scrolls while Screen Beaver records.
Polish + export
Zooms, captions, and voiceover land automatically. Export H.264 locally.
Agents in the wild[3/7]Cursor and Claude call Screen Beaver over MCP
Same protocol in both editors — ask for a demo, watch tools fire, and see the studio polish the take while you stay hands-off.
CursorMCPYou
Record a pricing-page demo with Screen Beaver. Add zooms + captions, then prepare export.
Tool · screenbeaver.record_and_build
{
"target": "pricing",
"captions": true,
"auto_zoom": true
}Result
Capture complete · 14 zooms · captions ready · export queued
Watching Screen Beaver UI update live…
ClaudeMCPUse live_build_demo — walk signup → dashboard → settings. Narrate with Edge TTS when done.
Calling screenbeaver.live_build_demo
- ✓ open_target dashboard
- ✓ click / type / scroll
- ✓ plant zooms + captions
- → generate_voiceover…
Studio is live. Review the timeline, then hit Export for local H.264 MP4.
Install[4/7]
One npm command. Same protocol everywhere.
Add Screen Beaver MCP globally, point Cursor or Claude Desktop at the server, and tools show up like native agent capabilities.
- Works in Cursor and Claude Desktop
- No browser extension required for MCP
- Desktop app unlocks native capture + Android USB
screenbeaver-mcp
# get Screen Beaver MCP
$ npm i -g @screenbeaver/mcp
# Cursor / Claude mcp.json
{
"mcpServers": {
"screenbeaver": {
"command": "screenbeaver-mcp"
}
}
}Agent toolkit
[5/7]What agents can call
Capture, edit, voice, and export — the same inspectors you see in the studio, exposed as MCP tools.
record_and_build
Capture a scripted walkthrough and run the auto-edit pipeline in one agent call.
live_build_demo
Drive the UI live — plant zooms, captions, effects, and voiceover while you watch.
Captures & targets
Pick screen, window, or app surfaces without fighting the OS share picker on desktop.
Edit & voice tools
Regenerate zooms, translate captions, speak with Edge TTS, pick music, prepare export.
The connector[6/7]
Plug agents into Screen Beaver.
MCP is the cable. Cursor or Claude clicks through your product while Screen Beaver records, polishes, and prepares the MP4 — tutorials without touching your screen.
Captions · Voice · Locale
Agents can translate and speak the demo too.
After the walkthrough, MCP tools can regenerate captions, translate locales, and run Edge neural TTS — so one take becomes many languages without re-recording.
Explore the full studio →FAQ · [7/7]
MCP FAQ
What is MCP?
Model Context Protocol — a standard way for agents like Cursor and Claude to call tools. Screen Beaver exposes capture, edit, and export tools over MCP.
Do I need the desktop app?
Full agent control (native capture, system-wide cursor, Android USB) needs the Windows desktop app. The web studio still works for manual recording.
Can I still record manually?
Yes. MCP is the unique path. The same studio works when you click it yourself — zooms, captions, voiceover, local export.
Make product videos your agent films for you
Start free, connect MCP when you're ready, and ship tutorials without touching your screen.