CloudGlue MCP Server
@aviaryhq
Official MCP server for Cloudglue API
Overview
What is CloudGlue MCP Server?
CloudGlue MCP Server connects Cloudglue to AI assistants like Cursor and Claude Desktop, enabling video collection understanding by turning videos into structured data ready for LLMs. It is designed for users who want to analyze videos programmatically through the Model Context Protocol.
How to use CloudGlue MCP Server?
Install via npx or download the Claude Desktop Extension, then configure with an API key from cloudglue.dev. Pass the API key using the --api-key argument in your MCP client JSON configuration. Once set up, use the available tools (e.g., list_collections, describe_video) to interact with Cloudglue.
Key features of CloudGlue MCP Server
- Connects Cloudglue to Cursor, Claude Desktop, and other AI assistants
- Unlocks video collection understanding for LLMs
- Supports YouTube, public HTTP, and data connector URLs (Dropbox, Google Drive, Zoom)
- Offers tools for discovery, individual video analysis, and collection analysis
- Provides page-based pagination across all major tools
- Includes intelligent cost optimization and automatic duplicate detection
Use cases of CloudGlue MCP Server
- Browse and search video collections and their metadata
- Get transcripts and descriptions from individual videos
- Extract structured entities using custom prompts
- Segment videos into camera shots or chapters
- Search across collections for specific moments or summaries
FAQ from CloudGlue MCP Server
What do I need to use CloudGlue MCP Server?
You need a Cloudglue API Key (obtain from cloudglue.dev) and an MCP client such as Cursor or Claude Desktop.
What video sources are supported?
Cloudglue files, YouTube URLs, public HTTP video URLs, and data connector URLs (Dropbox, Google Drive, Zoom). Some tools (e.g., segment_video_camera_shots) do not support YouTube URLs.
How does pagination work?
All tools use page-based pagination with a page parameter. Responses include page and total_pages fields. For describe_video, pagination is in 5‑minute segments with optional start_time_seconds.
Are there duplicate processing protections?
Yes, tools automatically check for existing transcripts, extractions, or segmentations before creating new ones to avoid redundant costs.
Where is my data processed and stored?
Data is processed and stored on Cloudglue’s platform. See the Privacy Policy and Terms of Service for details.