MCP WebSearch Server

MCP Server for web search and advanced web scraping. Uses SearchAPI.io for Google Search and AI Mode.

Quick Start

$3

``json
{
"mcpServers": {
"websearch": {
"command": "npx",
"args": ["-y", "mcp-websearch-server"],
"env": {
"SEARCHAPI_KEY": "your-api-key-here"
}
}
}
}
`

$3

Run as HTTP server:
`bash
MCP_TRANSPORT=http PORT=3000 npx mcp-websearch-server
`

Then configure MCP client with API key in URL or header:

Option A: API key in URL query parameter
`json
{
"mcpServers": {
"websearch": {
"url": "http://localhost:3000/mcp?apikey=YOUR_KEY",
"transport": "streamable-http"
}
}
}
`

Option B: API key in header
`json
{
"mcpServers": {
"websearch": {
"url": "http://localhost:3000/mcp",
"transport": "streamable-http",
"headers": {
"X-API-Key": "YOUR_KEY"
}
}
}
}
`

Option C: Server-side default key
`bash
MCP_TRANSPORT=http SEARCHAPI_KEY=default-key PORT=3000 npx mcp-websearch-server
`

The server implements MCP Streamable HTTP transport (spec 2025-03-26) with:
- Session management with Mcp-Session-Id headers
- SSE streaming for server-initiated messages
- Resumability support with Last-Event-ID
- Per-user API key support via query param or header

Get your API key at searchapi.io

Tools

| Tool | Description |
|------|-------------|
| web_search | Google Search with pagination, time filter, site filter |
| ai_search | Google AI Mode - get AI-generated answers with sources |
| web_scrape | Advanced scraper with multiple extract modes (text/markdown/structured) |
| get_links | Extract links from a webpage with optional filter |
| scrape_multiple | Scrape up to 5 URLs at once |

Parameters

$3

- query - Search query
- num_results - Number of results (1-20, default 10)
- page - Page number for pagination (1-10)
- time_period - Filter: last_hour, last_day, last_week, last_month, last_year
- site - Limit to specific site (e.g., "github.com")

$3

- query - Question or topic
- image_url - Optional image URL for visual questions
- location - Location for local queries

$3

- url - URL to scrape
- selector - CSS selector (optional)
- extract_mode - text, markdown, or structured
- include_links - Include links in output
- max_length - Max content length (1000-50000)

$3

- url - URL to extract links from
- filter - Text filter for URLs/anchors

$3

- urls - Array of URLs (max 5)
- max_per_page - Max content per page (500-5000)

Deployment

$3

1. Edit ecosystem.config.js and set your SEARCHAPI_KEY
2. Run: npm run pm2:start
3. Check logs: npm run pm2:logs

Or manually:
`bash
MCP_TRANSPORT=http PORT=3000 SEARCHAPI_KEY=your-key pm2 start index.js --name mcp-websearch
`

$3

`dockerfile
FROM node:18-alpine
WORKDIR /app
COPY package*.json ./
RUN npm ci --production
COPY . .
ENV MCP_TRANSPORT=http PORT=3000
EXPOSE 3000
CMD ["node", "index.js"]
``

Roadmap

- [ ] Add Serper.dev provider support
- [ ] Add SerpAPI provider support
- [ ] Provider auto-fallback

License

MIT