Upload and Manage Media Assets at Scale With Cloudinary MCP

If you manage media assets for a growing product, you know the challenge. Hundreds of new images, videos, and files arrive every week. Each one needs to be uploaded, organized, tagged, and optimized before it can go live. Doing this manually leads to duplicate assets, inconsistent tags, and files scattered across folders.

Cloudinary MCP (Model Context Protocol) connects your Cloudinary account to an AI agent like Claude. Instead of writing scripts or clicking through dashboards, you tell Claude what to do in plain English. Upload files, apply transformations, organize by metadata, generate delivery URLs. Claude calls the right Cloudinary tools and handles the execution.

In this guide, you’ll set up the Cloudinary Asset Management MCP server, connect it to Claude Desktop, and use natural language to automate uploading, transforming, organizing, and analyzing your media assets.

A working setup where Claude can:

• Upload images and videos to your Cloudinary Media Library from URLs or your local filesystem.
• Tag assets automatically at upload with categories, campaigns, and product metadata.
• Generate optimized variants (web, mobile, thumbnail) from a single source asset.
• Analyze assets with AI to generate auto-tags based on image content.
• Create asset reports showing file names, formats, and storage usage.

Cloudinary provides several MCP servers, each designed for a different part of the media workflow:

This guide focuses on the Asset Management MCP server, which handles uploading, transforming, and organizing assets. This is the foundation for any automated media workflow.

Before you start, make sure you have:

• A free Cloudinary account. You need your cloud name, API key, and API secret from the Cloudinary console (howto find them).
• Node.js version 18 or higher installed.
• Claude Desktop installed and running.

Open Claude Desktop, navigate to Settings, and click the Developers tab. You’ll see an option to add local MCP servers.

Click Edit Config. This opens a folder containing claude_desktop_config.json. You’ll add your Cloudinary MCP configuration to this file.

Open claude_desktop_config.json in your code editor and add the Asset Management MCP server:

Replace api_key, api_secret, and cloud_name with your actual credentials from the Cloudinary console. Save the file.

If your assets are stored locally (on your Desktop or Downloads folder), add the Filesystem MCP server alongside the Cloudinary one so Claude can access your local files:

Replace YOUR_USERNAME with your actual OS username. Add or remove directories in the argsarray based on where your files are stored.

Quit and reopen Claude Desktop so it loads the new configuration. Return to the Developers tab to confirm the MCP server is listed and running.

Test the connection by asking Claude:

If everything is configured correctly, Claude lists the available tools, including upload-asset, transform-asset, search-assets, and get-asset-details.

The connection is active. You can start managing assets.

With Claude connected to Cloudinary, you manage your media operations through conversation. You describe what you want done, and Claude calls the right tools.

When your product images are hosted online (from a photographer’s shared link or a staging CDN), you can upload and tag them in a single prompt:

Claude calls the upload-asset tool for each file, applies your tags, and returns the results with file names, formats, file sizes, and Cloudinary URLs.

The uploaded images are now stored in your Cloudinary Media Library, tagged and ready to use.

If your assets are stored locally and you’ve added the Filesystem MCP server (from Step 2), Claude can browse your local directories and upload files to Cloudinary.

First, you can ask Claude to organize your local files:

Then upload the entire folder:

Claude reads your local directory, passes each file through the upload-assettool, and applies your tags automatically.

This workflow is useful for large batches of media files from photo shoots, video productions, or audio recordings that originate on your machine.

Once assets are uploaded, Claude can generate optimized versions for different platforms. Ask for a single transformation:

Claude first calls search-assets to find everything tagged spring-2025, then calls transform-asset on each result with the parameters c_limit,w_1200/f_auto/q_auto.

You can also generate multiple variants in one prompt:

Claude applies each transformation and organizes the results in your Cloudinary folders.

If you have the Google Auto Tagging add-on enabled in your Cloudinary account, Claude can analyze assets and generate content-based tags automatically:

Claude calls the analysis tools to examine each file and enrich it with metadata like color, product type, or scene. These tags make your library searchable without manual categorization.

You can also organize assets by naming patterns:

When you need visibility into your library for audits, team handoffs, or tracking growth, ask Claude to generate a report:

Claude fetches the details from Cloudinary and returns a summary in chat. You can also ask for a .json or .csv export.

Here’s a walkthrough of the complete local file workflow. Files from a local “cloudinary” folder are uploaded to Cloudinary and organized into a “beach-2025” folder:

You connected Claude to Cloudinary through the Asset Management MCP server and used natural language to upload, tag, transform, analyze, and organize media assets. Every operation that would normally require manual dashboard work or custom scripts runs through a single conversational interface.

Here are a few ways to extend your setup:

• Add the Analysis MCP server. Connect the Cloudinary Analysis MCP server for AI-powered content recognition, moderation, and auto-tagging directly through Claude.
• Add the Structured Metadata MCP server. Create custom metadata fields (product line, region, expiration date) and apply them to assets through Claude for more advanced organization.
• Connect MediaFlows MCP. Build automated workflows that trigger on upload, metadata changes, or tag additions. See the MediaFlows MCP documentation for setup details.
• Combine with other MCP servers. If your team uses Google Drive, Dropbox, or Slack MCP servers, Claude can pull assets from those sources and upload them to Cloudinary in a single conversation.
• Build batch processing scripts. Use Claude to generate Node.js scripts that batch-process hundreds of assets with consistent tags, transformations, and folder structures.

For a deeper walkthrough of remote vs. local server setup and troubleshooting tips, read our blog post on automating media workflows with Claude and Cloudinary MCP.

Resources:

• Cloudinary MCP servers overview
• Asset Management MCP server
• Cloudinary upload documentation
• Image transformations documentation
• Google Auto Tagging add-on
• Filesystem MCP server

Cloudinary MCP is a set of Model Context Protocol servers that expose Cloudinary’s upload, transformation, analysis, and asset management capabilities as tools an AI assistant like Claude can call directly, instead of you writing API calls or scripts by hand. The full list of available servers and what each one covers is documented on the Cloudinary AI Agent Tools and MCP Servers page.

The Asset Management server is the foundation for most workflows, since it handles uploading, searching, and transforming media. From there you can add the Analysis server for AI-powered tagging and moderation, the Structured Metadata server for custom fields, or MediaFlows for automated triggers. The Cloudinary MCP servers GitHub repository lists all five servers and their npm packages.

No, not necessarily. Cloudinary offers both remote and local MCP servers. Remote servers are hosted by Cloudinary and support OAuth authentication, so you can connect without running anything on your own machine, while local servers run via npx and require your API credentials in a config file. Setup instructions for both approaches are in the MCP Servers documentation.

Yes, if you also connect the Filesystem MCP server alongside the Cloudinary Asset Management server. This lets Claude browse local directories, read files, and pass them to Cloudinary’s upload tool in the same conversation, the same pattern used in Cloudinary’s own guide to automating media workflows with Claude and Cloudinary MCP.

Yes. Rather than configuring each MCP server individually, Cloudinary offers a single marketplace plugin for Claude that installs all five servers, covering asset management, environment configuration, structured metadata, analysis, and MediaFlows, together. It’s available on the Cloudinary plugin page on Claude.com.