aidemo — Claude Code MCP server: what it does, how to install
Turns a one-line request into a narrated, captioned MP4: the agent writes storyboard.json, the engine drives Chrome, voices the scenes and cuts the result.
claude mcp add aidemo -- npx -y @tandryukha/aidemoWhat aidemo does
The server hands your agent a tool for filming product demos. You describe the flow in a sentence, and the agent writes storyboard.json: scenes, narration text, plus a fixed list of browser actions. The engine then opens a real Chrome window, types at a human pace and zooms into elements when the storyboard asks.
Rendering runs in stages. Voice first, from OpenAI, ElevenLabs or a local model. Then the browser replay, then word-level Whisper captions, then a compose pass that trims idle time, syncs audio to picture and writes output/final-demo.mp4. The same take also gives you a GIF and named stills for the README. Replays are deterministic, so when the UI shifts you re-run the committed storyboard; nobody re-records by hand.
An API key is needed for cloud voices only. Set AIDEMO_TTS_PROVIDER=local and everything renders offline, with nothing uploaded. Recording uses your own Chrome, which is why localhost and apps behind a login work fine. In CI one step, uses: tandryukha/aidemo@stable, replays the storyboard and commits fresh media.
What it helps with
- capture a 45-second checkout walkthrough without screen-studio editing
- refresh a demo after a redesign using the same committed storyboard
- add narration and synced captions to a screen recording
- export a README GIF and stills from one take
- keep demo media current from a CI job
Example prompts for the agent
- Film a 45-second walkthrough of checkout — login, cart, payment — and voice it in English.
- Write a
storyboard.jsonfor the onboarding tour, then render it locally without an API key. - Re-render
output/final-demo.mp4from the existing storyboard and pull stills for the README.
How to install aidemo
Platform CLI
claude mcp add aidemo -- npx -y @tandryukha/aidemoPlatform CLI
codex mcp add aidemo -- npx -y @tandryukha/aidemoConfig file
{
"mcpServers": {
"aidemo": {
"command": "npx",
"args": [
"-y",
"@tandryukha/aidemo"
]
}
}
}Config file
{
"servers": {
"aidemo": {
"type": "stdio",
"command": "npx",
"args": [
"-y",
"@tandryukha/aidemo"
]
}
}
}Platform CLI
gemini mcp add aidemo npx -y @tandryukha/aidemoStep by step, with commands for each platform: How to add an MCP server to Claude Code, Codex, Cursor, VS Code, Gemini CLI and more
Questions about aidemo
How do I install aidemo in Claude Code?
Run claude mcp add aidemo -- npx -y @tandryukha/aidemo. That registers the server, and the agent can then author storyboards and start a render.
What does aidemo do?
It produces a finished MP4 from a written storyboard: browser replay in Chrome, voiceover, captions and auto-zoom, trimmed and muxed into one file.
Do I need an OpenAI key?
Only if you use a cloud voice. With AIDEMO_TTS_PROVIDER=local the in-process model handles narration and the render stays offline.
