Cloudinary Test

You have an MCP server that implements the Model Context Protocol (MCP) API, enabling MCP clients to connect to Cloudinary test services and exchange model context data and actions. This guide shows how to use the server with an MCP client, how to install or run it locally, and how to configure it for various clients and environments.

How to use

Connect your MCP client to the remote server endpoint and include the required authentication headers provided by the server configuration. Use the MCP URL exposed by the server to access the MCP API under the /mcp path. When your client starts a session, it will negotiate model context, requests, and responses according to the MCP specification. If you already have a client configured for MCP, point it at the server URL and supply the headers shown in the configuration example to authorize requests.

How to install

Prerequisites you need before installation:

• Node.js installed (recommended)
• npm or npx available in your shell

Install and run the MCP server using the recommended local runtime. Run this command to start the server with your credentials and required options:

npx mcp start --api-key ... --api-secret ... --oauth2 ... --cloud-name ...

Additional configuration options

If you need to connect to the MCP server from an MCP client using a remote endpoint, you can use the following HTTP-based configuration that points to the server and includes the necessary authentication headers.

{
  "mcpServers": {
    "SDK": {
      "type": "mcp",
      "url": "https://cloudinary-test-mcp-server.kanwar-a2e.workers.dev/mcp",
      "headers": {
        "x-cloudinary-auth": "${MCP_CLOUDINARY_AUTH}",
        "x-oauth2": "${MCP_OAUTH2}"
      }
    }
  }
}

This server configuration for local and remote use

The server can be accessed in two common ways: as a remote MCP endpoint via HTTP or as a local process via standard input/output (stdio). The HTTP method uses a remote URL with authentication headers. The stdio method runs the MCP server locally using a command like the npm/npx start flow shown above. Choose the method that fits your deployment and client setup.

Security considerations

Protect your API credentials and OAuth tokens. Do not expose your api-key, api-secret, oauth2 tokens, or cloud-name in client configurations that could be read by unauthorized users. Use environment variables and secret management where possible, and rotate credentials according to your organization’s security practices.

Troubleshooting

If your MCP client cannot connect, verify the following: the endpoint URL is reachable from the client network, the required headers are present, and the credentials are correct. Check that the server is running locally or that the remote endpoint is live, and review any error messages from the MCP client for hints about authentication or network issues.

Notes on packaging and deployment

Deployment can be done as a remote HTTP MCP endpoint or as a local stdio-based server. For remote use, ensure the endpoint URL is publicly reachable and that the host allows MCP traffic. For local development, install dependencies, build if necessary, and start the server with the appropriate command and arguments.