MCP Integration

Prerequisites

Before starting with the ImgMCP platform, please ensure you have the following ready:

1. An ImgMCP Account: You need to register on the platform. If you don't have an account yet, please go to the Dashboard to complete the registration.
2. An API Key: An API key is required to authenticate your requests. After registering and logging in, you can find your default API key on the API Keys page in the Management Console.

Please keep your API key secure and confidential.

Core Configuration

• MCP Endpoint URL:

https://api.imgmcp.com/mcp

• Authentication Method: ImgMCP uses API keys for authentication. You need to provide the key in the Authorization header of your HTTP request in the following format:

Authorization: Bearer <Your API Key>

• Tip: If the MCP Host you are using makes it inconvenient to configure the Bearer <key> format, you can try using your API key directly as the value for the Authorization header. However, using the standard format including Bearer is recommended whenever possible.

MCP Host Configuration Guide

Cherry Studio

Cherry Studio supports the Streamable HTTP MCP protocol.

• Open Cherry Studio, navigate to Settings -> MCP Servers.
• Click Add Server, then select Import from JSON.
• Paste the following JSON content:

{
    "name": "ImgMCP",
    "type": "streamableHttp",
    "description": "Connecting Models to Your Creativity.",
    "provider": "ImgMCP",
    "providerUrl": "https://imgmcp.com",
    "logoUrl": "https://cdn.imgmcp.com/imgmcp-favicon.png",
    "baseUrl": "https://api.imgmcp.com/mcp"
}

• After importing, click on the newly created "ImgMCP" service in the list.
• In the Request Headers section, add your API key:

Authorization=Bearer <Your API Key>

(Make sure to replace <Your API Key> with your actual key)

• Click Save.

Chatwise

ChatWise supports direct configuration of Streamable HTTP MCP services.

• Open Chatwise, navigate to Settings -> Tools (or equivalent).
• Click the plus (+) icon in the bottom left to add a tool, select Streamable HTTP Request (or similar option).
• Enter the URL as https://api.imgmcp.com/mcp.
• Enter the HTTP Request Header as Authorization=Bearer <Your API Key>.
• Click View Tools (or equivalent). If configured correctly, the list of tools should appear below.

Claude Desktop

Claude Desktop does not currently support the Streamable HTTP MCP protocol directly. You need to use the mcp-remote tool to proxy requests.

• Open Claude Desktop, navigate to Settings -> Developer.
• Click the Edit Config button, which will open the claude_desktop_config.json file.
• Paste the following JSON content and save the file:

{
  "mcpServers": {
    "ImgMCP": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://api.imgmcp.com/mcp",
        "--header",
        "Authorization:${AUTH_HEADER}"
      ],
      "env": {
        "AUTH_HEADER": "Bearer <Your API Key>"
      }
    }
  },
  "globalShortcut": ""
}

(Replace <Your API Key> with your actual key)

• Restart Claude Desktop. ImgMCP should now be loaded.
• Note: By default, Claude Desktop does not display image links returned by MCP tools directly; you need to click the link to view the image.

Cursor

Cursor does not currently support the Streamable HTTP MCP protocol directly. You need to use the mcp-remote tool to proxy requests.

• Open Cursor, navigate to Settings -> MCP Tools (or equivalent AI/Tool settings).
• Click New MCP Server, which will open the mcp.json file (or prompt for configuration).
• Paste the following JSON content and save the file:

{
  "mcpServers": {
    "ImgMCP": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://api.imgmcp.com/mcp",
        "--header",
        "Authorization:${AUTH_HEADER}"
      ],
      "env": {
        "AUTH_HEADER": "Bearer <Your API Key>"
      }
    }
  }
}

(Replace <Your API Key> with your actual key)

• Return to the MCP Tools page (or refresh the settings view). If configured correctly, ImgMCP should be loaded.
• Note: By default, Cursor does not display image links returned by MCP tools directly; you need to click the link to view the image.

Verify Connection

After configuration, you need to verify that the MCP Host can successfully communicate with the ImgMCP platform using the new settings.

• Basic Check: In your MCP Host, check the list of available MCP tools. If tools from ImgMCP are displayed correctly, the basic connection is established.

• Functional Verification:

  1. Select an appropriate LLM within the Host and enable the ImgMCP toolset.
  2. Initiate a simple interaction. For example, try asking: "Query task history" or instruct it to perform an action that requires fetching information from the ImgMCP backend (like querying the list of available models).
  3. If the LLM responds correctly and returns information from ImgMCP (e.g., task history list or model list), the configuration is successful, and you can start using it normally.

Next Steps:

You have successfully configured your MCP Host. You can now explore the various AI model interaction capabilities provided by ImgMCP through your host. Refer to subsequent sections for more detailed usage examples.