Use Screenify from Claude Desktop, Cursor, and other AI apps via the Model Context Protocol.

The screenify CLI ships a built-in MCP (Model Context Protocol) server, so AI apps that can't run terminal commands — Claude Desktop, Cursor, ChatGPT's desktop developer mode, Windsurf and any other MCP host — can record your screen and produce studio-styled videos by talking to Screenify directly.

Using a coding agent that already runs shell commands (Claude Code, Codex)? You don't need MCP — those agents drive the CLI directly.

Zero-command setup (in the app)

The easiest path needs no terminal at all: open Screenify Studio → Settings → AI Assistants and click Connect AI assistants. It finds every AI client on your Mac and registers Screenify with each one — you'll see a per-app result right in Settings.

Then, in the Claude app: restart it, open a chat, and switch Screenify Studio on once in the chat input's + → Connectors menu — newly added connectors start toggled off, and the AI has no access to Screenify's tools until that switch is on.

One-command setup

Prefer the terminal? With the CLI enabled on your Mac:

screenify mcp install

This registers the server with every AI client detected on your machine (Claude Code, Claude Desktop, Cursor) and backs up any config it touches. Restart the client — and in Claude Desktop, remember the one-time Connectors toggle described above. To target one client, preview first, or check what's already registered:

screenify mcp install --target cursor
screenify mcp install --dry-run
screenify mcp status

Any other MCP host can launch the server manually — command screenify, args ["mcp"], stdio transport.

What the AI can do

Tool	What it does
`check_readiness`	Verify permissions before the first capture
`list_targets`	Screens, windows, cameras, mics, booted iOS Simulators
`record_clip`	One-shot: record for N seconds, style it (wallpaper, transitions, …) and even add cinematic effects — all in a single step
`start_recording` / `stop_recording`	Record a screen, window, several windows composited, or an iOS Simulator
`export_video`	Render a recording with studio styling — wallpaper, padding, device frames, transitions, aspect presets
`record_web_demo`	Drive a real browser over a URL (scripted or AI-driven) and produce a polished, auto-zoomed demo
`apply_effects`	Add cinematic zoom, spotlight and callout effects to a recording
`get_job_status` / `cancel_job`	Long renders run as background jobs the AI polls
`inspect_project` / `list_recordings` / `get_schemas`	Explore your library and the styling catalog
`open_result`	Open the finished video — or the recorded project in the Screenify Studio editor — right from the chat

A typical conversation in Claude Desktop: "record my main screen for 20 seconds, then export it vertical for TikTok with an orange wallpaper and a fade-in" — the AI chains the tools above and hands you back a finished MP4.

Permissions

Screen capture uses macOS screen-recording permission from the app that hosts the AI — the first time you record from Claude Desktop or Cursor, macOS will ask you to grant Screen Recording to that app (System Settings → Privacy & Security). check_readiness walks the AI through exactly what's missing. Exporting, effects and iOS Simulator recording work without it.

MCP server

Zero-command setup (in the app)

One-command setup

What the AI can do

Permissions

On this page