home / mcp / image recognition mcp server

Image Recognition MCP Server

Provides AI-powered image recognition and description capabilities via OpenAI-compatible vision models through the MCP protocol.

Installation

Add the following to your MCP client configuration file.

Configuration

View docs

{
  "mcpServers": {
    "akirose-image-recognition-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "@akirose/image-recognition-mcp"
      ],
      "env": {
        "OPENAI_MODEL": "gpt-5-mini",
        "OPENAI_API_KEY": "your-actual-openai-api-key-here",
        "ALLOWED_DOMAINS": "example.com,cdn.example.org",
        "ALLOW_ALL_PATHS": "true",
        "OPENAI_BASE_URL": "http://127.0.0.1:1234/v1",
        "ALLOWED_IMAGE_PATHS": "./images,./assets"
      }
    }
  }
}

You can add image analysis capabilities to your AI workflows with an MCP server that describes images using OpenAI-compatible vision models. This server lets assistants analyze images from URLs or local paths and returns detailed natural language descriptions, all through the MCP protocol for easy integration.

How to use

To use the image recognition MCP server, connect from an MCP client and invoke the describe-image tool to analyze an image from a URL or a local file path. You can prompt the server to tailor the description or ask specific questions about the image to get more targeted results. This enables your AI assistants to understand visual content and respond with rich, descriptive details.

How to install

Prerequisites you need before starting:

Node.js 18+
npm or yarn
OpenAI API key or a local vision model server (such as LM Studio or Ollama)

Install and run the MCP server using the provided MCP client configuration. The server runs as a standalone MCP service that you deploy locally or in your environment.

{
  "mcpServers": {
    "image_recog": {
      "command": "npx",
      "args": ["-y", "@akirose/image-recognition-mcp"],
      "env": {
        "OPENAI_API_KEY": "your-actual-openai-api-key-here"
      }
    }
  }
}
```

```
{
  "mcpServers": {
    "image_recog_all": {
      "command": "npx",
      "args": ["-y", "@akirose/image-recognition-mcp"],
      "env": {
        "OPENAI_API_KEY": "your-actual-openai-api-key-here",
        "ALLOW_ALL_PATHS": "true"
      }
    }
  }
}

Additional setup notes

Important environment variables you may configure to customize behavior include the base URL for vision models, the model name, allowed image paths, and domain restrictions for security. You should provide a valid OPENAI_API_KEY and, if using local models, set OPENAI_BASE_URL to point at your local server.

Security and usage notes

The server includes path validation, extension checks for image files, optional domain whitelisting, and file existence checks to help prevent unauthorized access or invalid inputs. When ALLOW_ALL_PATHS is enabled, the server still restricts inputs to common image file extensions for security.

Troubleshooting

If you run into issues, verify that your OPENAI_API_KEY is set and valid. Check that the key is accessible to the MCP server and that the target vision model is reachable (local or OpenAI). Review any network or permission errors reported by the MCP runtime.

Notes on integration

This MCP server is designed to integrate with AI assistants that support the MCP protocol. You can call describe-image to analyze an image and receive a structured, descriptive response suitable for feeding into chat flows or decision making.

Available tools

describe-image

Analyzes an image from a URL or local file path and returns a detailed description. You can supply an optional prompt to tailor the description.