# MCP Search & Fetch

An MCP server exposing Ollama's web search and web fetch capabilities as tools.

## Features

- **web_search**: Perform web searches using Ollama's hosted search API
- **web_fetch**: Fetch content from web pages

## Requirements

- Ollama API key ([get one from ollama.com](https://ollama.com/login))

## Installation

```bash
pip install mcp-search-and-fetch

# Export your API key
OLLAMA_API_KEY=your_api_key_here

# Run the MCP server
mcp-search-and-fetch
```

### Clone from Repository

```bash
# Clone the repository
git clone https://github.com/lkiesow/mcp-search-and-fetch.git
cd mcp-search-and-fetch

# Install dependencies
pip install -r requirements.txt

export OLLAMA_API_KEY=your_api_key_here
python mcp_search_and_fetch.py
```

### Docker

To run the pre-built containers:

```bash
docker run -p 8000:8000 --env-file .env ghcr.io/lkiesow/mcp-search-and-fetch:latest
```

Or, if you want to build your own containers:

```bash
docker build -t mcp-search-fetch .
docker run -p 8000:8000 --env-file .env mcp-search-fetch
```

### Docker Compose

An example ´docker-compose.yml` with Caddy as reverse proxy:

```yml
services:
  search-and-fetch:
    image: ghcr.io/lkiesow/mcp-search-and-fetch:1.0.1
    container_name: search-and-fetch
    restart: always
    environment:
      OLLAMA_API_KEY: secret.ollama.api.key
    networks:
      - mcp

  caddy:
    image: docker.io/library/caddy:2.10.2
    container_name: caddy
    restart: always
    environment:
      CADDY_DOMAIN: search-and-fetch.example.com
      CADDY_API_KEY: secret.mcp.api.key
    volumes:
      - /opt/mcp-search-and-fetch/caddy:/etc/caddy
      - caddy_data:/data
      - caddy_config:/config
    ports:
      - 80:80
      - 443:443
      - 443:443/udp
    networks:
      - mcp

volumes:
  caddy_data:
  caddy_config:

networks:
  mcp:
```

And an example `Caddyfile` in the `caddy` directory like this:

```Caddyfile
{$CADDY_DOMAIN} {
	@no_auth {
		not header Authorization "Bearer {$CADDY_API_KEY}"
	}
	respond @no_auth "Unauthorized" 401
	reverse_proxy /* search-and-fetch:8000
}
```

## Usage

### Local (stdio)

```bash
python mcp_search_and_fetch.py
```

### HTTP Server (Streamable HTTP)

```bash
# Set port and host in .env
export MCP_SERVER_PORT=8000
export MCP_SERVER_HOST=0.0.0.0

python mcp_search_and_fetch.py
```

## Configuration

You can either set environment variables or provide a `.env` file.
Take a look at the [.env.sample](.env.sample) for an example.

| Environment Variable | Required | Default   | Description
|----------------------|----------|-----------|------------------------------
| `OLLAMA_API_KEY`     | Yes      | -         | Your Ollama API key
| `MCP_SERVER_PORT`    | No       | -         | Run HTTP server on specified port
| `MCP_SERVER_HOST`    | No       | 127.0.0.1 | Host to bind to

## Tools

### web_search(query, max_results=3)

Performs a web search and returns results as JSON.

### web_fetch(url)

Fetches content from a URL and returns it as JSON.