Developers and builders can now connect Cloudinary to their preferred LLM.
The new Cloudinary Model Context Protocol (MCP) (beta) server is officially live! Now you can connect AI agents to Cloudinary functions to test rapidly, automate more creatively, and develop cross-system workflows without writing code. (See full press release here.)
Along with new OpenAPI specs and AI-ready documentation, the release of our MCP server is the next step toward a new era of building where developers, vibe coders, and non-technical builders can create and launch rich visual apps, products, product MVPs, demos, and more using natural language.
For those unfamiliar with Model Context Protocol, here’s a quick explanation.
LLMs like Claude are powerful for generating text, but they don’t have access to live data, external tools, or real-time context. Matt Noyes, Cloudinary’s Senior Director of Product Marketing, sums it up this way: “MCP, as a set of rules, fills that gap. It gives these models a way to use tools and data they weren’t trained on and can’t access by themselves.”
With the help of MCP, these LLMs and AI agents can go beyond their traditional limitations to use outside tools and handle complex requests without being constantly retrained.
For example, after allowing Claude access to Cloudinary’s MCP server, you can tell it, “Find all assets with the tag ‘holiday2025’ and move them to the folder ‘holiday2025_campaign’.” Claude will complete that task on your behalf, then respond with: “Done! I’ve added the tag ‘holiday2024’ to all images in the current folder.”
Cloudinary users can now manage, transform, and optimize visual media with natural language prompts.
Here’s just some of what you can expect:
- LLM-readable and retrievable documentation. Access Cloudinary docs as structured Markdown, with chunk-level metadata and code blocks optimized for AI tools.
- Accurate, updated API calls. Forget outdated answers or hallucinated API calls. Tools like Cursor and Claude now fetch official docs directly from Cloudinary’s Markdown and MCP sources.
- Dedicated documentation and video tutorials. Learn from tutorials covering MCP configuration in various IDEs, tips, and use cases for getting the most out of Cloudinary’s MCP server.
- Real-time API awareness. Automate creatively with Claude and OpenAI Assistants that can now (via API):
- Understand and navigate the API structure.
- Validate input types and required parameters.
- Generate accurate SDK-based calls (Node.js, Python, etc.).
- Suggest inline examples directly in the IDE.
- Exposed upload and Admin APIs via dedicated MCP servers. Give LLM tools programmatic access to Cloudinary endpoint definitions and parameters.
- Execution-ready APIs. By providing an API key and endpoint, empower AI tools to begin chaining or calling Cloudinary API operations in live tasks or CLI scripts.
- Build using Base44 integration. Base44 is a no-code automation agent with a Cloudinary integration with preloaded knowledge of Cloudinary’s product and capabilities. Users of any skill level can quickly create full apps using natural language prompts.
You can use your AI agent of choice to tag files, create metadata fields, crop assets, automate workflows, and complete a host of other tasks. Here are just a few ideas to get you started:
- Upload assets. Claude can receive a file and upload it to a specific folder or location in Cloudinary.
- Organize media. Tag, categorize, or move assets based on metadata.
- Delete or archive. Automatically remove outdated or unused media as part of a cleanup.
- Create multi-step automations. Chain taste like uploading a file, tagging it, applying a transformation, generating delivery URLs.
- Integrate across tools. Trigger or respond to events from other systems using an AI agent as a coordinator.
- Search media assets. Retrieve files using natural language prompts like: “Retrieve all image assets with the tag ‘car’”.
- Uncover media and insights. Use Claude to uncover assets based on metadata and tags, and ask for suggestions on how to improve.
With the release of our new MCP server, we hope to empower builders of all kinds to explore the future of creating rich visual experiences with Cloudinary. And we’re just getting started. These tools pave the way for a suite of Cloudinary-native agents that will automate content workflows, moderation, transformations, and delivery, putting creative power in the hands of every builder.
For more detailed information, visit Cloudinary Documentation.
