Image Analyzer

This Image Analyzer MCP Server runs locally alongside Ollama to provide image analysis, OCR, UI analysis, and batch processing directly within MCP-enabled apps like Claude Desktop, without requiring any API keys. It emphasizes privacy by keeping all data on your device and works offline once set up.

How to use

You use the Image Analyzer MCP Server by configuring an MCP client to connect to the local Ollama-backed service. After setup, you can issue prompts in your MCP client to analyze images, extract text, analyze UI layouts, or process multiple images at once. The server stays on your device, so your data does not leave your machine.

How to install

Prerequisites you need before starting:

Install Ollama on your system. You can install via the following options depending on your platform.

# Windows: From the official download site, install the Windows package

# macOS
brew install ollama

# Linux
curl -fsSL https://ollama.ai/install.sh | sh

Start the Ollama service and download the image analysis model (about 7GB).

ollama serve
ollama pull gemma3:12b

Configure your MCP client to connect to the local server. Use the following example as a template and replace the placeholder with your actual project path.

{
  "mcpServers": {
    "image-analyzer": {
      "command": "python",
      "args": ["你的项目路径/image_analyzer_server.py"],
      "env": {
        "OLLAMA_BASE_URL": "http://localhost:11434",
        "OLLAMA_MODEL": "gemma3:12b"
      }
    }
  }
}

Save this configuration in your MCP client configuration file. For example, on Windows you might place it in %APPDATA%\Claude\claude_desktop_config.json, and on macOS in ~/Library/Application Support/Claude/claude_desktop_config.json. Then restart the MCP client to apply changes.

Additional configuration and notes

Environment and model configuration are centralized in the MCP config. The Ollama base URL and the model name must match your local Ollama setup.

If you need to adjust timeouts or log levels, you can extend the environment configuration in your client setup or use standard environment variables as needed by your workflow.

Troubleshooting

If the tool doesn’t appear in your MCP client, verify the following: Ollama is running, the configuration file path is correct, and you have restarted the client after changes.

Common issues include the model not being downloaded or Ollama port contention. Ensure the gemma3:12b model is downloaded and that port 11434 is free.

Security and privacy

All processing runs locally. No data is uploaded or sent to external servers, and you do not need an API key.

Available tools

analyze_image

Performs basic image content analysis to understand what's in the picture.

analyze_image_categories

Analyzes and categorizes UI elements and visual design aspects for UX/UI analysis.

extract_text_from_image

Extracts all readable text from an image using OCR.

batch_analyze_images

Processes multiple images in a single run to return aggregated results.

check_ollama_status

Checks the status of the Ollama service to ensure it is running correctly.

list_supported_formats

Lists image formats supported by the service.