Cloudinary Media Generation Server

Summary

Image Generation API: Use the Image Generation API to generate images from text prompts using various AI models.

The API supports Basic Authentication using your Cloudinary API Key and API Secret, which can be found on the API Keys page of your Cloudinary Console.

Key Features:

• Unified API: A single interface for generating images across multiple best-in-class AI models.
• Cloudinary Integration: Generated images are automatically available for delivery, transformation, and optimization through Cloudinary's platform.
• Future-proof: Adopt new state-of-the-art models as they become available, without rebuilding your integration.

Supported Model Families:

• flux: Photorealistic images (FLUX.2 Klein 9B / FLUX.2 Pro).
• recraft: Vector and illustration (Recraft V3 / Recraft V4).
• gpt-image: Campaign and marketing images (GPT Image 1 Mini / GPT Image 2).
• nano-banana: General purpose generation (Nano Banana 1 / Nano Banana 2).
• ideogram: Realism, text rendering, and artistic generation (Ideogram V4).

The Image Generation API requires the Cloudinary Image Generation add-on.

Learn more

Note:

This is an early version of our Image Generation API. As the capability grows, certain features and endpoints may be adjusted. We invite you to try it out and share your feedback with our support team.

Table of Contents

• Cloudinary Media Generation MCP Server
  • Installation
  • Custom server / non-default host
  • Progressive Discovery
  • Development
  • Contributions

Installation

{
  "command": "npx",
  "args": [
    "@cloudinary/media-generation-mcp",
    "start",
    "--api-key",
    "",
    "--api-secret",
    "",
    "--cloud-name",
    ""
  ]
}

claude mcp add CloudinaryMediaGeneration -- npx -y @cloudinary/media-generation-mcp start --api-key  --api-secret  --cloud-name 

gemini mcp add CloudinaryMediaGeneration -- npx -y @cloudinary/media-generation-mcp start --api-key  --api-secret  --cloud-name 

{
  "command": "npx",
  "args": [
    "@cloudinary/media-generation-mcp",
    "start",
    "--api-key",
    "",
    "--api-secret",
    "",
    "--cloud-name",
    ""
  ]
}

{
  "command": "npx",
  "args": [
    "@cloudinary/media-generation-mcp",
    "start",
    "--api-key",
    "",
    "--api-secret",
    "",
    "--cloud-name",
    ""
  ]
}

npx @cloudinary/media-generation-mcp start --api-key  --api-secret  --cloud-name 

npx @cloudinary/media-generation-mcp --help

Custom server / non-default host

By default the server talks to the production Cloudinary API (https://api.cloudinary.com/v2). To point it at a different host — staging, a regional endpoint, or a local mock — pass --server-url (or for clients that use a config block, add it to args):

npx @cloudinary/media-generation-mcp start \
  --api-key  --api-secret  --cloud-name  \
  --server-url https://api-eu.cloudinary.com/v2

In a client config block:

{
  "command": "npx",
  "args": [
    "@cloudinary/media-generation-mcp",
    "start",
    "--api-key", "",
    "--api-secret", "",
    "--cloud-name", "",
    "--server-url", "https://api-eu.cloudinary.com/v2"
  ]
}

Notes:

• Keep the /v2 suffix. Operation paths (/processing/{cloud_name}/...) are appended to this base, so a host without /v2 will return 404s.
• --server-url overrides the URL entirely. --server-index selects from the schema's servers list, which currently has a single entry, so only --server-index 0 is valid — use --server-url for anything else.
• {cloud_name} is independent of the host. It is always taken from --cloud-name / CLOUDINARY_CLOUD_NAME / CLOUDINARY_URL, regardless of --server-url.
• Add --log-level debug to print the outgoing request URL and confirm the override took effect.

Progressive Discovery

MCP servers with many tools can bloat LLM context windows, leading to increased token usage and tool confusion. Dynamic mode solves this by exposing only a small set of meta-tools that let agents progressively discover and invoke tools on demand.

To enable dynamic mode, pass the --mode dynamic flag when starting your server:

{
  "mcpServers": {
    "CloudinaryMediaGeneration": {
      "command": "npx",
      "args": ["@cloudinary/media-generation-mcp", "start", "--mode", "dynamic"],
      // ... other server arguments
    }
  }
}

In dynamic mode, the server registers only the following meta-tools instead of every individual tool:

• list_tools: Lists all available tools with their names and descriptions.
• describe_tool_input: Returns the input schema for one or more tools by name.
• execute_tool: Executes a tool by name with its arguments.

This approach significantly reduces the number of tokens sent to the LLM on each request, which is especially useful for servers with a large number of tools.

Development

Run locally without a published npm package:

1. Clone this repository
2. Run npm install
3. Run npm run build
4. Run node ./bin/mcp-server.js start --api-key --api-secret --cloud-name

To use this local version with Cursor, Claude or other MCP Clients, you'll need to add the following config:

{
  "command": "node",
  "args": [
    "./bin/mcp-server.js",
    "start",
    "--api-key",
    "",
    "--api-secret",
    "",
    "--cloud-name",
    ""
  ]
}

Or to debug the MCP server locally, use the official MCP Inspector:

npx @modelcontextprotocol/inspector node ./bin/mcp-server.js start --api-key  --api-secret  --cloud-name 

Contributions

While we value contributions to this MCP Server, the code is generated programmatically. Any manual changes added to internal files will be overwritten on the next generation. We look forward to hearing your feedback. Feel free to open a PR or an issue with a proof of concept and we'll do our best to include it in a future release.

MCP Server Created by Speakeasy