Taboola API MCP Server

A flexible MCP (Model Context Protocol) server with fetchRecommendations functionality. Supports both local (STDIO) and remote (HTTP) deployment modes.

Setup

1. Install dependencies:

pip install -r requirements.txt

2. Activate virtual environment (if using one):

source .venv/bin/activate

Related MCP server: tavily-search-mcp-server

Deployment Options

Local Mode (STDIO Transport)

Perfect for local development and testing with MCP Inspector:

# Default mode - runs locally with STDIO transport
python server.py

# Explicitly specify local mode
python server.py --mode local

Remote Mode (HTTP Server)

Deploy as a remote HTTP server accessible over the network:

# Run as HTTP server on default port 8000
python server.py --mode remote

# Specify custom host and port
python server.py --mode remote --host 0.0.0.0 --port 3000

# Using environment variables
export MCP_MODE=remote
export MCP_HOST=0.0.0.0
export MCP_PORT=8000
python server.py

Configuration Options

Command Line Arguments

• --mode: Server mode (local or remote) - default: local

• --host: Host to bind to in remote mode - default: 0.0.0.0

• --port: Port to bind to in remote mode - default: 8000

Environment Variables

• MCP_MODE: Server mode (local or remote)

• MCP_HOST: Host to bind to in remote mode

• MCP_PORT: Port to bind to in remote mode

Environment variables override command line arguments.

Functions

fetchRecommendations

Fetches recommendations for a given publisher using their API key via Taboola API.

Parameters:

• publisher_name (str): The name of the publisher

• api_key (str): The API key for authentication

Returns:

• str: JSON recommendations data from Taboola API

Usage Examples

Local Development with MCP Inspector

# Start server locally
python server.py

# In another terminal, run MCP Inspector
npx @modelcontextprotocol/inspector python server.py

Remote Deployment

# Deploy as remote server
python server.py --mode remote --port 8000

# Server will be available at: http://your-server-ip:8000
# Connect using HTTP transport with MCP clients

Production Deployment

For production, consider using environment variables:

export MCP_MODE=remote
export MCP_HOST=0.0.0.0
export MCP_PORT=8000
python server.py

Or with a process manager like PM2:

pm2 start server.py --name "taboola-mcp" -- --mode remote --port 8000

Testing

Use the provided test script to verify functionality:

# Edit test_function.py with your credentials
python test_function.py

Cloud Deployment

Render Deployment

Deploy easily on Render cloud platform:

Option 1: Using Render.yaml (Recommended)

1. Push your code to GitHub/GitLab

2. Connect to Render:

   • Go to Render Dashboard

   • Click "New" > "Blueprint"

   • Connect your repository

   • The render.yaml file will be automatically detected

3. Deploy:

   • Render will automatically build and deploy your MCP server

   • Your server will be available at: https://your-app-name.onrender.com

Option 2: Manual Render Setup

1. Create a new Web Service on Render

2. Connect your repository

3. Configure the service:

   • Build Command: pip install -r requirements.txt

   • Start Command: python server.py --mode remote --host 0.0.0.0 --port $PORT

   • Environment Variables:

     • MCP_MODE=remote

     • MCP_HOST=0.0.0.0

     • PYTHON_VERSION=3.13.0

4. Deploy and get your URL

Docker Deployment

For any Docker-compatible platform:

# Build and run locally
docker build -t taboola-mcp-server .
docker run -p 8000:8000 taboola-mcp-server

# Or use docker-compose
docker-compose up -d

Other Cloud Platforms

The server is compatible with:

• Heroku: Use Procfile with web: python server.py --mode remote --port $PORT

• Railway: Deploy directly from GitHub with automatic detection

• DigitalOcean App Platform: Use the provided docker-compose.yml

• AWS/GCP/Azure: Deploy using Docker or direct Python deployment

Security Notes

• In remote mode, the server binds to 0.0.0.0 by default (all interfaces)

• Consider using a reverse proxy (nginx, Apache) for production deployments

• Ensure proper firewall rules are in place for remote access

• API keys are passed as parameters - ensure secure transmission (HTTPS recommended)

• Cloud platforms like Render automatically provide HTTPS endpoints