Connect via MCP

Operate your Moviie library and AI features from an AI assistant using the Model Context Protocol (MCP).

Moviie exposes a Model Context Protocol (MCP) server so AI assistants can operate your account by natural language: manage videos, generate AI learning assets, read analytics, and check credits, all scoped to your organization.

It works with any MCP client that connects to a remote server over streamable HTTP with an Authorization header — including Claude Code, Claude Desktop, Cursor, Windsurf, and VS Code. Per-client setup is below.

The server is a small, curated set of opinionated tools, not a one-to-one mirror of the REST API. Each tool is designed so the assistant picks the right one and knows how to use it.

Endpoint

https://mcp.moviie.ai/mcp

Authentication

The MCP server authenticates with a private API key (the same mvi_ keys you use for the REST API), passed as a Bearer token. Publishable keys are rejected, because MCP tools can create, update, and delete content.

Open your dashboard and go to Settings → API Keys.
Create a private key and copy it (it is shown only once).
Give it to your MCP client as shown for your client below.

Keep the key server-side

Treat the key like a password. Where your client supports it, reference the key through an environment variable (e.g. ${env:MOVIIE_API_KEY}) instead of pasting it into a shared config file. If a key leaks, delete it from Settings → API Keys and issue a new one.

Set up your client

Claude Code

Add the server from the terminal:

claude mcp add --transport http moviie https://mcp.moviie.ai/mcp \
  --header "Authorization: Bearer mvi_your_private_api_key"

Cursor

Add it to ~/.cursor/mcp.json (global) or .cursor/mcp.json (per project):

{
  "mcpServers": {
    "moviie": {
      "url": "https://mcp.moviie.ai/mcp",
      "headers": {
        "Authorization": "Bearer mvi_your_private_api_key"
      }
    }
  }
}

Windsurf

Add it to mcp_config.json (Cascade → Manage MCP servers → View raw config). Note the field is serverUrl:

{
  "mcpServers": {
    "moviie": {
      "serverUrl": "https://mcp.moviie.ai/mcp",
      "headers": {
        "Authorization": "Bearer mvi_your_private_api_key"
      }
    }
  }
}

VS Code

Add it to .vscode/mcp.json (workspace) or your user mcp.json. VS Code uses the servers key with "type": "http":

{
  "servers": {
    "moviie": {
      "type": "http",
      "url": "https://mcp.moviie.ai/mcp",
      "headers": {
        "Authorization": "Bearer mvi_your_private_api_key"
      }
    }
  }
}

Claude Desktop

Claude Desktop connects to remote servers through the mcp-remote bridge. Add this to claude_desktop_config.json (Settings → Developer → Edit Config):

{
  "mcpServers": {
    "moviie": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://mcp.moviie.ai/mcp",
        "--header",
        "Authorization:${MOVIIE_AUTH}"
      ],
      "env": {
        "MOVIIE_AUTH": "Bearer mvi_your_private_api_key"
      }
    }
  }
}

Why the header is split

The Authorization value is passed via the MOVIIE_AUTH env var (with no space after the colon in args) to sidestep a known bug where some desktop clients mangle spaces inside args. Env-var values keep their spaces, so Bearer … arrives intact.

Any other client

Any client that supports a remote streamable-HTTP MCP server works: point it at https://mcp.moviie.ai/mcp and set an Authorization: Bearer mvi_… header. If your client only speaks stdio (no header field), use the mcp-remote bridge shown for Claude Desktop.

Once connected, ask the assistant to list your videos to confirm it works.

What the assistant can do

The tools are grouped by intent:

Area	What you can ask
Library	List videos, get a video's details, rename or recategorize, delete, list collections
Search	Find a phrase inside a video's transcript
AI generation	Generate a summary, chapters, a mind map, an ebook, or an interactive quiz from a video
Learning pack	Generate a mind map, ebook, and quiz together in one step
Dubbing & subtitles	Dub a video into another language, or translate its subtitles
Status & retrieval	Check whether a generation is done, then fetch the finished asset
Player & embed	Get ready-to-paste embed code, or manage call-to-action buttons
Analytics	Read plays, watch time, retention, and geography for a video or the whole account
Credits	Check your credit balance, and preview what a generation will cost before running it

How generations work

A transcript is required. AI generations run from the video's transcript. Transcription happens automatically after upload; if it is not ready yet, the tool says so and no credits are charged. Try again shortly.
Generations are asynchronous. The assistant triggers the work, then polls for status and retrieves the finished asset. Text assets are usually ready in a minute or two; dubbing and translation take longer and scale with video length.
Costs are transparent. You can preview a generation's cost and check your balance before committing. A generation you cannot afford is not started.
Destructive actions confirm first. Deleting a video (or removing a call-to-action) is a two-step action: the assistant previews exactly what will change, then applies it only after you confirm.

Rate limits

MCP requests count against the same per-key and per-organization limits as the REST API. See Rate Limits for the budgets and how throttling is signaled.

Connect via MCP

On this page