How to Make a Product Demo Video From the Command Line

The DemoVexa product demo video CLI lets you make a demo from the command line: install DemoVexa with npm, start it with one command, and ask Claude Code to make the video for you. The terminal handles install, start, updates and logs, while the studio does the scripting, voiceover and animation from your app's code. The result is a narrated MP4 in 16:9, 9:16 and 1:1, made on your own computer.

#Why would a developer use a product demo video CLI?

Because the terminal is already open. If you just merged a feature, switching to a separate app and clicking through a wizard is the step that gets skipped. A command you can run in the same window, or a request you can type to Claude Code, has a better chance of happening on release day.

A CLI also fits machines where you'd rather not install a desktop app, such as a Linux box you reach over SSH or a laptop where you keep tools in npm. The DemoVexa docs describe the npm version as "the same studio" as the desktop app, on macOS, Windows and Linux.

There is one thing to be clear about. This is not a headless renderer you script line by line. The CLI starts and manages DemoVexa. The video itself is made in the studio, either by you or by Claude acting on your behalf.

#How do you install the DemoVexa CLI?

Install it from npm, then start it. You need Node.js 20 or newer.

npm install -g demovexa
demovexa run

The first run downloads the current build (about 350 MB), sets up the video engine and opens DemoVexa in your browser at http://local.demovexa.com:1557. That address points at your own computer. From there the setup matches the desktop app: sign in, connect Claude and ElevenLabs, and add your product as a Workspace. The install guide covers each step.

[Image: a terminal running demovexa run next to the studio open in a browser]

#Useful options for demovexa run

• --port 1600 runs it on another port.
• --no-open starts it without opening the browser.
• --ip uses 127.0.0.1 instead of local.demovexa.com, for browsers that insist on HTTPS for that name.

#Keep it separate from the desktop app

The CLI keeps its data in its own folder: ~/Library/Application Support/DemoVexa CLI on macOS, %APPDATA%\DemoVexa CLI on Windows and ~/.local/share/demovexa-cli on Linux. Set DEMOVEXA_HOME to move it. Because the folders differ, you can have the CLI and the desktop app installed side by side.

#Which commands does the CLI include?

The CLI has nine commands, all for running and maintaining the studio:

demovexa logs --follow is the one to keep open in a second terminal while a build runs. If something fails, the reason is there.

The local address uses plain http, which browsers mark "Not secure" even though nothing leaves your machine. demovexa trust creates a small certificate authority on your computer and adds it to the system trust store, so the browser shows a lock. The key stays in the data folder.

#How do you make the video from your terminal with Claude Code?

Connect Claude Code to DemoVexa once, then ask for videos in plain words. Claude talks to DemoVexa through MCP, the Model Context Protocol, and DemoVexa makes the video as if you had clicked the buttons.

#Connect it

In DemoVexa, open App settings, then Connect to Claude. Under Claude Code, copy the command. It starts with claude mcp add. Paste it into your terminal and start a new Claude Code session. The demovexa mcp command also prints the connection command.

[Image: App settings with the Connect to Claude page and the Claude Code command]

#Ask for the video

Right after you ship, ask something like "make a launch video for what we just built". According to the docs, Claude Code passes a summary of the change, the changed files and the commit messages to DemoVexa, so the video shows the right screens. Other requests the docs list as examples:

• "What's the status of my latest video?"
• "Render the vertical version."
• "Give me the LinkedIn post for Data Rooms."
• "Suggest three video ideas for the Contacts module."

Making a video this way takes it from brief to finished output: script, voiceover, scene plan, build, renders, posts and covers. You can ask for a length, up to 3 minutes. See the Claude MCP page for the full list of tools.

#What happens between the command and the MP4?

The same steps as in the app. DemoVexa writes a Brief from your request, then a Script, scene by scene. ElevenLabs reads it as the Voice in any of 85 languages, or you can skip ElevenLabs and use captions over music. In Animate, a build agent recreates your screens as React components from your code and animates them with Remotion, using clean sample data. Export renders 1080p MP4s in 16:9, 9:16 and 1:1, and the Marketing kit adds covers and post copy.

Nothing here records your screen. That matters for developers, because the next time the UI changes you rebuild instead of re-recording. The features page walks through each step.

[Image: the scene plan for a launch video with each scene tied to a source file]

#Where your code goes

DemoVexa runs on your computer. Your code, voiceovers, projects and renders stay on your disk. Only what a build needs goes straight to Claude through your own Anthropic account, and API keys sit in your operating system's keychain. The DemoVexa cloud receives sign-in, licence and usage counts, not your code.

#What can't the CLI do yet?

Based on the current docs, the CLI does not have commands like "make video" or "render" that run without the studio. Those steps happen in the studio or through Claude over MCP. If you want several videos built unattended, the studio's batch builds can run overnight, but you start them from the Projects page, not from a flag.

If you are planning a CI job that renders a demo on every merge, check the docs for changes first. Today the honest answer is that Claude Code in your terminal is the closest thing to a scripted workflow.