@aviaryhq/cloudglue-mcp-server NPM

Cloudglue MCP Server

Connect Cloudglue to Cursor, Claude Desktop, and other AI assistants to unlock the power of video collection understanding. Cloudglue helps turn your videos into structured data ready for LLMs.

📖 Resources

By using the Cloudglue SDK and/or the MCP server, you agree to the Cloudglue Terms of Service and acknowledge our Privacy Policy.

Setup

1. API Key

First, get a Cloudglue API Key from cloudglue.dev, this will be used to authenticate the MCP server with your Cloudglue account.

2. Configure MCP client

Next, configure your MCP client (e.g. Claude Desktop) to use this MCP server. Most MCP clients store the configuration as JSON, for cloudglue-mcp-server this would look like the following:

{
  "mcpServers": {
    "cloudglue": {
      "command": "npx",
      "args": [
        "-y",
        "@aviaryhq/cloudglue-mcp-server@latest"
      ],
      "env": {
        "CLOUDGLUE_API_KEY": "<YOUR-API-KEY>"
        }
    }
  }
}

Replace <YOUR-API-KEY> with the API Key created in step 1.

Optional Configuration

You can customize the server with additional CLI arguments:

--api-key: Provide API key directly (alternative to environment variable)
--base-url: Use custom Cloudglue API endpoint
--working-dir: Set working directory for local file uploads (defaults to home directory)

Example with custom working directory:

{
  "mcpServers": {
    "cloudglue": {
      "command": "npx",
      "args": [
        "-y",
        "@aviaryhq/cloudglue-mcp-server@latest",
        "--working-dir",
        "/path/to/your/videos"
      ],
      "env": {
        "CLOUDGLUE_API_KEY": "<YOUR-API-KEY>"
        }
    }
  }
}

Local Development Setup

1. API Key

First, get a Cloudglue API Key from cloudglue.dev, this will be used to authenticate the MCP server with your Cloudglue account.

2. Install and build server locally

You can build this server locally via:

npm install
npm run build

3. Configure MCP client

Next, configure your MCP client (such as Cursor) to use this server. Most MCP clients store the configuration as JSON in the following format:

{
  "mcpServers": {
      "cloudglue-mcp-server": {
          "command": "node",
          "args": [
              "/ABSOLUTE/PATH/TO/PARENT/FOLDER/cloudglue-mcp-server/build/index.js",
              "--api-key",
              "<YOUR-API-KEY>"
          ]
      }
  }
}

Tools

The following Cloudglue tools are available to LLMs through this MCP server:

File Management

add_file: Upload local files to Cloudglue or add existing Cloudglue files to collections. Supports two modes: 1) Upload new file with optional collection assignment, 2) Add existing file to collection. Returns comprehensive file metadata and collection status. Use working directory context for relative file paths. Supports videos, images, documents, and other file types with automatic MIME type detection.

Collection Management

create_collection: Create a new Cloudglue collection for organizing videos with specific analysis configurations. Supports both rich-transcripts (comprehensive video analysis) and entities (structured data extraction) collection types. Each type has different configuration options for customizing the analysis pipeline. Returns complete collection metadata upon successful creation.
add_youtube: Add YouTube videos to a Cloudglue collection with parallel processing. Supports individual video URLs (up to 50), playlist URLs, and channel URLs. For playlists/channels, extracts up to 15 most recent videos using YouTube RSS feeds. Note: Only YouTube videos with available transcripts can be successfully processed by Cloudglue. Processes videos in batches of 5 for optimal performance and waits for all to complete before returning the updated collection video list.

Discovery & Navigation

list_collections: Discover available video collections and their basic metadata. Use this first to understand what video collections exist before using other collection-specific tools. Shows collection IDs needed for other tools, video counts, and collection types.
list_videos: Browse and search video metadata with powerful filtering options. Use this to explore available videos, find specific content by date, or see what's in a collection. Returns essential video info like duration, filename, and IDs needed for other tools.

Individual Video Analysis

get_video_description: Get comprehensive transcripts and descriptions from individual videos (YouTube or Cloudglue upload) with intelligent cost optimization. Automatically checks for existing transcripts before creating new ones. Supports customizable summarization.
get_video_entities: Extract structured data and entities from videos using custom prompts with intelligent cost optimization. Automatically checks for existing extractions before creating new ones. For individual videos - use retrieve_collection_entities for bulk collection analysis. The quality of results depends heavily on your prompt specificity.
get_video_metadata: Get comprehensive technical metadata about a Cloudglue video file including duration, resolution, file size, processing status, and computed statistics. Use this when you need video specifications, file details, or processing information rather than content analysis. Different from content-focused tools like get_video_description.

Collection Analysis

retrieve_collection_transcripts: Bulk retrieve rich multimodal transcripts (text, audio, and visual) from a collection with advanced filtering. Use this for comprehensive analysis of multiple videos in a collection, when you need to compare transcripts, or analyze patterns across content. For single videos, use get_video_description instead. Use date filtering to focus on specific time periods.
retrieve_collection_entities: Batch retrieve structured entity data from multiple videos in a collection. Entities can be user-defined based on what's important for your collection (people, objects, concepts, custom categories). Perfect for data mining, building datasets, or analyzing previously extracted entities at scale. Supports pagination and date-based filtering to manage large result sets. For individual video entities, use get_video_entities instead.
find_video_collection_moments: AI-powered semantic search to find specific moments, topics, or content within a video collection. Returns relevant segments with context, timestamps, and citations. Perfect for finding needle-in-haystack content, specific discussions, or thematic analysis across multiple videos. Much more targeted than bulk retrieval tools.

When to Use Which Tool

To create collections: Use create_collection to set up new video collections with specific analysis configurations
To upload files: Use add_file to upload local files or add existing files to collections
To add YouTube content: Use add_youtube to bulk add YouTube videos or playlists to collections
Start exploring: Use list_collections and list_videos to explore available content
For single videos: Use get_video_description or get_video_entities
For collections: Use retrieve_collection_* for bulk analysis or find_video_collection_moments for targeted searches
For technical specs: Use get_video_metadata