Cloudinary MCP Server

What is Cloudinary MCP Server?

Cloudinary MCP Server is a Model Context Protocol server that exposes Cloudinary Upload and Admin API methods as callable tools for AI assistants. It enables AI systems to upload and manage assets in a Cloudinary cloud on behalf of the user.

How to use Cloudinary MCP Server?

Install and configure by adding an entry to your MCP client’s configuration (e.g., claude_desktop_config.json) with the command npx -y cloudinary-mcp-server and environment variables CLOUDINARY_CLOUD_NAME, CLOUDINARY_API_KEY, and CLOUDINARY_API_SECRET. The server provides five tools: upload, delete-asset, get-asset, find-assets, and get-usage.

Key features of Cloudinary MCP Server

• Upload assets to Cloudinary via URL, file, or base64
• Delete assets by public ID or asset ID
• Retrieve detailed asset information including metadata
• Search assets with query expressions and pagination
• View Cloudinary usage report (storage, credits, bandwidth, etc.)

Use cases of Cloudinary MCP Server

• AI assistants uploading images or videos to a Cloudinary cloud
• Searching and retrieving existing assets by tags or folder paths
• Deleting unwanted or expired assets through natural language commands
• Checking Cloudinary usage and account limits programmatically

FAQ from Cloudinary MCP Server

What prerequisites are needed to run this server?

Node.js and an MCP client (such as the Claude Desktop App) are required. You also need a Cloudinary API key and secret, which you can create in the Cloudinary console.

How does the server authenticate with Cloudinary?

Authentication is handled via three environment variables: CLOUDINARY_CLOUD_NAME, CLOUDINARY_API_KEY, and CLOUDINARY_API_SECRET. These are passed in the MCP client configuration.

What data does the server access?

The server interacts with your Cloudinary cloud using the Cloudinary Admin API. All asset uploads and management actions occur within your own Cloudinary account.

Are all Cloudinary API methods supported?

No, only a subset of Upload and Admin API methods are currently available (upload, delete-asset, get-asset, find-assets, get-usage). More methods will be added over time. You can request new methods by opening an issue on the GitHub repository.

How is the server installed and run?

The server is installed and run automatically via npx (no global install needed). It communicates with the MCP client over stdio. No additional server runtime is required beyond Node.js.