How to record product demo videos with OpenClaw

TL;DR: Install the screencli skill from ClawHub, tell OpenClaw what to demo, and get a polished video with a shareable link. No screen recording app, no video editor, no retakes. The whole thing takes under two minutes.

The problem with recording demos manually

Recording a product demo the traditional way takes 90–120 minutes: launch a screen recorder, rehearse the flow, re-record when you click the wrong thing, trim the dead air, export, upload, share. According to Walnut’s 2026 demo report, even teams with templates spend 30–40 minutes per demo after initial setup.

If you’re already using OpenClaw to automate tasks across your apps and APIs, switching to a manual screen recorder feels like a step backward. You want to describe the demo, let the agent handle it, and move on. (For the full landscape of how AI agents create demo videos, we covered this in a previous post.)

What if your AI agent recorded the demo for you

OpenClaw already navigates browsers, manages files, and calls APIs through its skills system. With the screencli skill installed, it can open a browser, follow your instructions, and record the entire session as a polished video — all from a chat command.

Here’s what the workflow looks like:

1. You describe the demo in plain English
2. OpenClaw launches a browser and follows your prompt
3. screencli records the session with auto-zoom, click highlights, and cursor trails
4. The video uploads automatically and you get a shareable link

No screen recording app running in the background. No manual editing. No retakes.

Step 1: Install the screencli skill

Install from ClawHub with one command:

clawhub install ai-demo-recorder

This adds the screencli skill to your OpenClaw workspace. OpenClaw will discover it automatically when you ask it to record a demo. You can verify it’s installed with clawhub list.

Step 2: Record a public page

The simplest case — record a walkthrough of a public URL:

npx screencli record https://your-app.com -p "Navigate to the pricing page, compare the Free and Pro plans, then click Sign Up"

The AI agent opens the page, follows your instructions step by step, and records everything. When it finishes, you get:

✓ Video composed — recordings/a3f2c8e1/composed.mp4
✓ https://screencli.sh/v/a3f2c8e1
  1 credit used · 9 remaining

A composed MP4 with gradient background, auto-zoom on each action, click highlights, and a cursor trail. Plus a hosted link you can drop into a PR, Slack, or README.

Or just tell OpenClaw directly in chat:

Record a demo of https://your-app.com — show the pricing page and compare the plans.

OpenClaw runs the screencli command for you and sends back the shareable link.

Step 3: Record behind a login wall

Most product demos need authentication. Use --login to log in manually first, then hand off to the agent:

npx screencli record https://app.example.com -p "Show the dashboard metrics and export the monthly report" --login --auth myapp

The browser opens for you to log in. Once done, the AI takes over. Your login is never recorded. The --auth myapp flag saves the session so next time you can skip the manual login:

npx screencli record https://app.example.com -p "Walk through the settings page" --auth myapp

Step 4: Choose a background and export

screencli applies a gradient background automatically. Override it if you want a specific look:

npx screencli record https://your-app.com -p "Demo the checkout flow" --background sunset

Six gradients available: midnight, ember, forest, nebula, slate, copper.

Need a specific platform format? Export after recording:

npx screencli export ./recordings/a3f2c8e1 --preset twitter
npx screencli export ./recordings/a3f2c8e1 --preset gif

Why OpenClaw + screencli work well together

OpenClaw’s skill system makes screencli a natural fit. Here’s why:

One-command install. clawhub install ai-demo-recorder — no npm global installs, no config files, no path management. The skill lands in your workspace and OpenClaw picks it up.

Chat-driven workflow. Instead of switching to a terminal, you can ask OpenClaw to record directly from WhatsApp, Discord, or whatever chat platform you have connected. Describe the demo in natural language, get the link back in the same conversation.

Composable with other skills. OpenClaw’s 13,000+ skills on ClawHub mean you can chain workflows. Deploy with one skill, record the result with screencli, post the link to Slack with another — all from a single chat thread.

When to use this

PR reviews. Drop a 15-second GIF showing the UI change instead of describing it in text. --preset gif gives you a README-ready file.

Changelog posts. Record each new feature as you ship it. Takes one command, not a 30-minute recording session.

Onboarding docs. Record the setup flow once with --auth and the session is reusable for re-recordings when the UI changes.

Bug reports. Record the reproduction steps: npx screencli record https://staging.app.com -p "Click the export button three times rapidly". Attach the link to the ticket.

How it compares

Loom and Screen Studio require you to drive the browser manually. A manual screen recorder means you’re clicking through the flow yourself and editing out mistakes. screencli lets your AI agent handle the browser while it records, applies post-production effects automatically, and gives you a shareable link. For a comparison of the underlying browser automation frameworks, see our breakdown of browser-use, Stagehand, Playwright, and Skyvern.

FAQ

Do I need an Anthropic API key? No. screencli provides the AI through its cloud. Sign in with GitHub on first run and you get 15 free credits per month.

Can I trigger recordings from chat? Yes. With the screencli skill installed, just message OpenClaw: “Record a demo of the signup flow on staging.” It runs the command and sends back the link.

What if the agent does something wrong during recording? Refine your prompt and re-record. It takes under a minute. No editing — just re-run with a better description.

Does it work with apps behind auth? Yes. Use --login --auth <name> on the first run to log in manually. The agent takes over after login and reuses the saved session on future recordings.

What does the output look like? A polished MP4 with gradient background, rounded corners, drop shadow, auto-zoom to each action, click highlights, and smooth cursor trails. Plus a hosted shareable link at screencli.sh.

Install the skill → clawhub install ai-demo-recorder

Record your first demo → npx screencli record https://your-app.com -p "your demo prompt"