How to record automated UI walkthroughs with Claude Code and Screen Beaver

Automating the Software Demo Walkthrough

Recording software demos by hand gets tedious fast. Every minor UI adjustment requires another manual recording pass. You click around the screen, try to move your mouse smoothly, and hope you do not misclick. Then you open an editor to add zooms, captions, and crop out messy desktop borders. If you use terminal-based coding agents like Claude Code or IDE tools like Cursor, you already automate code generation. You can automate the demo video capture too.

Screen Beaver provides an optional npm package called @screenbeaver/mcp. This package implements the Model Context Protocol (MCP). It allows coding agents to control screen recording sessions directly. Instead of recording your monitor while clicking through a browser tab or local desktop app, you prompt your agent to execute the walkthrough. Screen Beaver handles the capture, applies visual enhancements like auto zoom and cursor smoothing, and exports a local MP4 video using H.264 encoding.

Prerequisites and Environment Setup

Before wiring up the agent integration, you need the core recording studio running on your system. Screen Beaver is a Windows screen recorder available on the Microsoft Store with a free tier and a Pro version. Install the desktop application and launch it on your Windows machine.

The desktop app must be active in the background for the MCP server to execute commands. Screen Beaver captures full displays, target windows, browser tabs, and Android screens over USB. Make sure your target application or local dev server is open and accessible before pointing an agent at it.

Connecting @screenbeaver/mcp to Your Agent

With the desktop app running, you install the MCP server integration. The package is distributed through npm. You can run it directly using npx inside your agent configuration file.

If you use Claude Code, Cursor, Windsurf, or VS Code, locate your MCP server configuration file (often named mcp.json). Add a new server entry for Screen Beaver:

{ "mcpServers": { "screenbeaver": { "command": "npx", "args": ["@screenbeaver/mcp"] } } }

This single npm package handles communication across compatible MCP clients, including GitHub Copilot and Claude Code. Once saved, your coding agent gains access to Screen Beaver tools alongside its standard file editing and terminal capabilities.

Running an Automated Walkthrough

Once the protocol connection is established, you do not need to click around your application to produce a demo. You instruct your agent in plain text. For example, you can tell Claude Code to launch a browser session, perform a specific user flow, and record the walkthrough.

The agent sends structured commands over MCP to initiate recording. Screen Beaver captures the target window or tab while the agent interacts with your application. Because the agent manages the interaction script, you eliminate human error during the capture process. There are no sudden pauses, accidental mouse drifts, or wrong tab selections.

Post-Capture Styling and Local Export

Raw screen captures usually look plain. Screen Beaver automatically turns raw recordings into polished demos immediately after the take finishes. The software processes the recording on your local machine using hardware-friendly H.264 encoding.

Automated edits include several built-in visual effects:

• Automatic Zoom: The editor tracks active focal points and zooms in dynamically where action occurs.
• Smooth Cursor Effects: Jerky cursor paths are smoothed, and click sounds or visual indicators are inserted.
• Captions: Text captions generate automatically along the timeline to highlight key steps.
• Background Blur and Framing: Wallpapers can be blurred or replaced with gradients and padded frames to hide background clutter.
• Cropping: Trim edges before framing your software on the canvas.

All processing happens locally on Windows. No raw video files or private app data get uploaded to third-party cloud servers just to encode an MP4 file. Your code and product visuals remain strictly on your local hardware.

Manual Studio Controls

Automated agent captures handle quick walkthroughs, but you retain full manual control over the editing timeline. After the agent completes a run, open the Screen Beaver window. You can adjust camera inserts, modify zoom levels, trim cuts, adjust background blur percentages, or edit caption text line by line.

If you prefer to record without an agent, the desktop studio operates as a standalone tool. You can hit record by hand, capture display or Android USB screens, plant manual zoom markers, adjust framing padding, and export locally. The MCP server provides hands-off capture when you want it, while the core app serves as your daily screen recording workflow.