Record and narrate demo, tutorial, onboarding, and product-walkthrough videos of your running web app — directly from Cursor or Claude Code. Cinematic zoom, smooth cursor animation, AI voiceover (or your own cloned voice). One prompt, no editor.
This plugin packages the narrateai-demomaker MCP server + its companion skill into a single bundle that works in both Cursor and Claude Code / Claude Desktop. It drives a headless browser through your app and ships a fully narrated MP4 via the NarrateAI pipeline.
You say:
"Make a demo of my checkout flow. Sign in as my test user (creds in my
.env.local)."
Cursor loads the bundled skill, reads your repo (routes, selectors, auth helpers, product copy), composes a structured plan + narration brief, asks you to confirm/skip the cinematic zoom steps, then drives an isolated browser session and ships a narrated MP4. End-to-end usually takes 1–3 minutes for a 30–90-second video.
Use cases the same one prompt handles:
- 🚀 Product demos — feature showcases for launch tweets, landing pages, sales follow-ups
- 🎓 Feature tutorials — "how to use X" videos that stay in sync with your live code
- 🚪 Onboarding walkthroughs — softer narration tone, automatically
- 📖 Doc-update videos — refresh embedded help-centre videos after a UI change
- 🐛 Bug repros — "here's exactly what happens when you click X"
- 📣 Launch teasers — drop a 30-sec narrated clip into a Product Hunt post
Browse the Cursor Marketplace, search "DemoMaker", click Install.
Browse the Claude plugin directory (or run /plugin > Discover inside Claude Code), find "NarrateAI DemoMaker", click Install. Or via the CLI:
/plugin install narrateai-demomaker@claude-plugins-officialRun this once in your terminal to mint your free API key — it auto-bakes the key into your IDE's MCP config:
uvx narrateai-demomaker initYour browser opens, you sign in with Google or GitHub, you get 5 free minutes of narrated video. No copy-pasting.
If you don't use Cursor or Claude:
curl -sSL https://narrateai.app/install.sh | bashThis installs uv (if missing), then runs uvx narrateai-demomaker init for you.
| Component | Path |
|---|---|
| MCP server entry | mcp.json |
| Skill (zoom-preview UX, brief composition, codebase-aware planning) | skills/narrateai-demomaker/SKILL.md |
| Reference docs (plan schema, iframe targeting, troubleshooting, etc.) | skills/narrateai-demomaker/references/ |
The MCP server itself is shipped as a Python package on PyPI and runs via uvx — it doesn't bundle into this plugin repo, so updates ship instantly without you needing to update the plugin.
- Free: 5 minutes of narrated video, no credit card. Free-tier videos include a small NarrateAI watermark.
- One-Time Package: $9.99 — 30 minutes, no expiration.
- Growth / Pro: monthly subscriptions with more minutes + watermark removal.
All plans share a single credit pool — your minutes work the same whether you're using the web app, the narrateai-mcp server, or DemoMaker. Manage at narrateai.app.
- The MCP server runs locally on your machine. It opens a headless browser that hits your own URLs only (localhost or your live deployment).
- Your API key is stored in Cursor's
mcp.json— same place every other MCP credential lives. - Recording is uploaded to NarrateAI's pipeline (necessary for narration). Videos are deleted after processing unless you explicitly save them in your library.
- Source code is never sent to NarrateAI — the agent reads your code locally to compose a plan; only the resulting plan + raw screen recording leaves your machine.
- 📚 Product page
- 📦 PyPI package (MCP server)
- 🐛 Report an issue
- 💬 Support
MIT — see LICENSE.