Link Scan

You can deploy and use the Link Scan MCP Server to automatically detect, analyze, and summarize links from video and text sources. It combines local AI summarization with lightweight transcription and HTML content extraction, delivering concise, 3-sentence Korean summaries without requiring API keys. This makes it easy to integrate into MCP workflows and client applications.

How to use

Use an MCP client to connect to the Link Scan MCP Server and request two main capabilities: scanning video links (YouTube, Instagram Reels) and scanning text links (blogs, articles). The server returns compact summaries that are suitable for quick consumption in dashboards, assistants, or automated workflows.

To test drive from a client, you can reference the explicit MCP command example that runs the local server and then issue requests to the corresponding tool endpoints. If you prefer a remote deployment, provide the public HTTP endpoint once it is running and reachable over HTTPS.

How to install

Prerequisites you need before installation include Python 3.11 or newer, ffmpeg for audio processing, Ollama for local LLM execution, and a compatible environment for container-based deployments.

Step-by-step setup you can follow locally:

# Clone the project
git clone https://github.com/your-username/mcp-link-scan.git
cd mcp-link-scan

# Install dependencies
pip install -r requirements.txt

System dependencies you should install according to your platform:

# ffmpeg for audio processing
# macOS
brew install ffmpeg

# Ubuntu/Debian
sudo apt-get install ffmpeg

# Windows: install from the official site

Ollama is required for local LLM operation. Install and pull the model before running the server:

# Install Ollama
# macOS
brew install ollama

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

# Windows: use the installer from Ollama
# Then pull the model
ollama pull llama3:latest

Run the server in local standard I/O mode or in HTTP mode as shown in the usage examples:

# Local stdio server
python -m src.server

# Remote HTTP server
python run_server.py

# Or run via uvicorn directly for HTTP
uvicorn src.server_http:app --host 0.0.0.0 --port 8000

Additional configuration and notes

Environment variables you may configure include port, host, debug mode, and Ollama-related settings. You can customize the API prefix for the MCP path if you host multiple MCP servers on the same host.

Important runtime notes:

• Audio/video dependencies

Security and deployment tips

Expose your server with HTTPS for production use and consider a firewall or gateway in front of the MCP endpoint. If you deploy to a cloud host, ensure the public URL is reachable by MCP clients.

For development, hot reload and local Ollama usage help you iterate quickly while keeping dependencies isolated.

Testing with MCP Inspector

You can validate server behavior using the MCP Inspector tool to view schemas, run sample inputs, and confirm protocol compliance.

Available tools

scan_video_link

Scan and summarize video links (YouTube, Instagram Reels, etc.).

scan_text_link

Scan and summarize text links (blogs, articles, etc.).