MCP

Clipelle for AI assistants

A small server that speaks the Model Context Protocol and calls the Clipelle API with your key. Any assistant that supports MCP over stdio can then create ads, translate videos and read projects for you, in a conversation.

What it is

The server runs on your own machine, reads your API key from the environment and exposes four tools. It never stores anything itself: every call goes to the API, every video lands in your Studio like any other project, with the same credits and limits.

A conversationtext
You: Make a 30-second ad for the Nimbus bottle. Photo: https://yourshop.com/images/nimbus-front.jpg
Assistant → create_ad({ name, productName, description, images: [{ url }], seconds: 30, quality: "standard" })
Assistant: Started project 7c0d…; I will check back.
Assistant → get_project({ id: "7c0d…" })
Assistant: The video is ready: <link, valid for about an hour>

Install

The server lives in the mcp/ folder of the Clipelle repository and is not published as a package. You need Node 20 or newer.

Buildbash
git clone https://github.com/YOUR-ORG/clipelle.git
cd clipelle/mcp
npm install
npm run build   # writes dist/index.js

Configure

Make a key on Settings → API keys, then add the server to your assistant's MCP configuration. The shape below is what most desktop assistants use; the key stays in the environment, never in a prompt.

MCP configurationjson
{
  "mcpServers": {
    "clipelle": {
      "command": "node",
      "args": ["/absolute/path/to/clipelle/mcp/dist/index.js"],
      "env": {
        "CLIPELLE_API_KEY": "rl_your_key",
        "CLIPELLE_API_URL": "https://YOUR-DOMAIN"
      }
    }
  }
}

CLIPELLE_API_URL defaults to the public site when left out. For a local build of the app, point it at your dev server.

Tools

  • create_ad: product name, description, public image links, length, quality, language and the optional choices; returns the project id and status.
  • translate_video: a public video link, its length in seconds, the target languages and the rights confirmation; returns the project.
  • get_project: status, steps and the result link of one project.
  • list_projects: your most recent projects.

Every tool returns the API's JSON, so an assistant can read the status and the plain-language error of a failed step.

Safety

  • Creating a video spends credits. Most assistants ask before running a tool; keep that on.
  • The server only uses files you link to; nothing on your machine is uploaded.
  • Revoke the key on Settings → API keys and the server stops working immediately.