Cloudinary

This connector bridges Cloudinary's Asset Management MCP server, Cloudinary's official remote MCP server at https://asset-management.mcp.cloudinary.com/mcp. Cloudinary's own server provides the tools: uploading from a URL or a data URI, listing, searching and updating assets, folders, tags and metadata, usage reports, AI image generation, and transformations. AnythingMCP puts them behind your MCP endpoint with OAuth for Claude and ChatGPT, per-tool roles, the audit log and response mapping, next to your other systems.

What it is for: many APIs only take a file by public URL. Instagram and Facebook publishing, for example, fetch the image or video from a link you give them. cloudinary_upload_asset takes a public URL, a Base64 data URI (an AI-generated image, a file from another tool) or a private S3 or Google Cloud Storage path, stores the file in your Cloudinary product environment and returns secure_url, an HTTPS link anyone can fetch. Pass that link on to the next connector.

The tools come from the server. Cloudinary signs you in before it serves anything, so the catalog's list is installed first; once you authorize, AnythingMCP asks the server for its tools and takes their current descriptions and schemas. Cloudinary adds tools over time: open the connector and click Discover tools to pick up new ones.

Setup (about 2 minutes):

1. Install the connector. There is nothing to paste.
2. On the connector page click Authorize with Provider. AnythingMCP reads Cloudinary's OAuth metadata, registers itself as a client (dynamic client registration) and sends you to Cloudinary.
3. Sign in, choose the product environment (cloud) the AI should use and approve. Back in AnythingMCP the tools are listed again from the server.

No API key, client id or secret is needed. If you would rather not sign in (a server-to-server setup), Cloudinary's server also accepts the API key: in the connector's settings switch authentication to API key, header name cloudinary-url, and paste the value of API environment variable from Settings > API Keys in the Cloudinary Console, which has the form cloudinary://<api_key>:<api_secret>@<cloud_name>.

Whose access it is: the person who authorizes becomes the Cloudinary user for every caller of this connector, in the product environment picked at sign-in. Everyone using it through AnythingMCP uploads, changes and reads that environment's assets. For a shared connector, authorize with a Cloudinary user whose role only covers what the AI should do. To switch to another product environment, click Authorize with Provider again and pick it.

Public by default: an asset uploaded with the default delivery type upload is public to anyone who has its URL. Do not upload confidential files unless you set type to authenticated or private, or use access_control; those URLs then need a signed link and are no use to an API that fetches by URL.

Safety: deleting assets (cloudinary_delete_asset), derived assets (cloudinary_delete_derived_assets), folders (cloudinary_delete_folder) and asset relations (cloudinary_delete_asset_relations) install switched off; turn them on per tool if needed. Uploads, renames, tag and metadata changes stay on. Renaming an asset (cloudinary_asset_rename) changes its public ID and so its URL: links already handed out stop working.

Working with it:

• Upload: cloudinary_upload_asset with upload_request.file set to the URL or data URI. Set asset_folder, public_id, tags or context as needed, and use the returned secure_url and asset_id. resource_type auto picks image, video or raw from the file.
• Find: cloudinary_search_assets with an expression such as tags=summer AND resource_type:image AND created_at>1w, or the cloudinary_list_images / _videos / _files lists. Results page with next_cursor.
• Read and change one asset: cloudinary_get_asset_details and cloudinary_asset_update take the immutable asset_id; the bulk tag, context and metadata tools take public IDs.
• Transformations: read cloudinary_get_tx_reference once per session, then build a delivery URL such as https://res.cloudinary.com/<cloud_name>/image/upload/c_fill,w_1080,h_1080/f_auto/q_auto/<public_id>, or create derived files with cloudinary_transform_asset. A transformed URL works as soon as it is requested; Instagram wants JPEG, so use f_jpg rather than f_auto for it.
• AI images: cloudinary_generate_image (text to image) and cloudinary_generate_image_from_images (from reference images) store the result as an asset with a public URL. They need the Image Generation add-on and use credits.
• Files on the user's own computer: cloudinary_sign_upload returns signed parameters for posting the file straight to Cloudinary; the AI cannot read local paths through AnythingMCP.
• cloudinary_get_usage_details shows credits, storage and bandwidth against the plan.

Other Cloudinary MCP servers: Cloudinary also runs servers for environment configuration (upload presets, named transformations, webhooks), structured metadata and media analysis. Add them as custom MCP connectors if you need them; this connector covers the asset library.