A local-first demo stack for NDA-bound software

Recording a demo of NDA-bound software creates a familiar problem: the video should be easy to review, but its screens may expose customer names, unreleased features, internal URLs, or credentials. A local-first workflow can reduce the number of places that footage travels. It cannot make an unreviewed recording safe by itself.

For a Windows team, the practical stack is simple: prepare a controlled app environment, capture and edit on a managed desktop, then export a local H.264 MP4. Screen Beaver is designed for Windows and can record full screens, specific windows, browser tabs, or Android screens over USB. It converts captures into demos with auto zoom, cursor effects, captions, and local H.264 export. There is no forced cloud upload for encoding.

That makes it a useful part of a privacy-first demo video workflow. It does not establish that the whole workstation, app, or optional agent connection meets an organization’s air-gap or NDA requirements. Treat those as separate checks.

1. Make the software environment boring

Start with an isolated test machine or approved Windows workstation, not a developer’s everyday desktop. Use a demo account and seeded data. Keep production credentials, customer records, internal chat, and unrelated browser tabs out of the environment. If the software is built from an offline code repository, prepare and verify the build and its dependencies before disconnecting the machine or moving it into a restricted network.

Offline source control helps limit exposure during development, but it does not sanitize the app’s interface. A test build can still show real endpoints, names, or secrets. Before recording, walk through the exact flow and replace sensitive values with representative data. Decide which screens may appear and which should not.

2. Capture only what the viewer needs

Choose a specific app window or browser tab when that is enough to tell the story. A full-screen capture can reveal notifications, taskbars, or other open applications. Window and tab capture reduce that risk, but they do not hide sensitive content inside the chosen surface. For practical framing advice, see how to record a clean app window without showing your desktop.

Screen Beaver captures microphone, camera, system audio, and cursor path tracks. Decide which tracks belong in the demo before recording. A spoken explanation can help a viewer follow a task, but it can also disclose names or details that never appear on screen. Keep the room quiet, silence notifications, and use a short test take to check both the picture and sound.

Crop, background blur, padding, and drop shadows can improve presentation. They are editing controls, not security controls. Blur may leave text partly readable, and a later crop can expose content outside the intended framing. Do not use either as a substitute for removing sensitive data at the source. Review the full frame and the exported file.

3. Keep capture and rendering on the workstation

Screen Beaver encodes video locally on the desktop and exports H.264 MP4, avoiding a required upload to a rendering service. That is a meaningful boundary: the footage need not leave the machine just to become a finished file. Local rendering also makes the workstation responsible for the processing and the resulting file, so confirm it has enough capacity and that the export destination is approved.

For stricter environments, have security or IT verify the workstation’s network policy and the complete software installation before use. Local export is not proof that every component makes no network connection, and the product facts do not establish compatibility with every air-gapped configuration. Treat the workflow as local-first until your organization has tested and approved the specific setup. The trade-offs of moving rendering onto the desktop are covered in this look at local rendering and the export pipeline.

4. Add agent control only when it fits policy

Screen Beaver offers an optional Model Context Protocol (MCP) agent connection through the npm package @screenbeaver/mcp. An MCP-capable coding agent can control walkthroughs and record videos automatically; the same studio can also be used manually. Automation can make repeatable steps less dependent on mouse timing, but it adds another component to review.

For an air-gapped workflow, do not assume the package or the coding agent can be installed or operated without network access. Have the team responsible for the environment approve how the package and agent are obtained, where they run, and what access they have. If those checks are not complete, record manually and leave the optional agent out of the stack.

5. Review the MP4 as a deliverable

After export, review the entire video at normal speed. Check captions, cursor movement, audio, transitions, and every screen that appears. Confirm the file is saved in the approved location, then share it through the organization’s approved channel. Keep the original capture only as long as policy requires.

This process trades the convenience of cloud rendering for desktop capacity, file handling, and more responsibility for the capture environment. It is a sensible route to local H.264 screen capture when a team wants to limit footage transfers. The strongest protection still comes before the recording starts: controlled data, a prepared screen, and a deliberate review.