Overview

Search1API’s MCP server exposes a single, unified MCP (Model Context Protocol) endpoint that brings web search, news discovery, crawling, and site sitemap extraction into a plug-and-play service for LLM assistants and developer tools. Instead of integrating multiple search providers and scraping logic yourself, you configure an MCP client to forward user queries to this server and receive structured, enriched results (including optional full-page crawl content).

This is useful for building assistants that need up-to-date web facts, news monitoring, content ingestion, or lightweight web crawling without managing infrastructure. The server supports both remote HTTP-hosted mode (Streamable HTTP MCP) and a local stdio mode (via npx), and accepts per-session API keys so you can scope access by user or workspace.

Features

Unified web search across many services (Google, Bing, DuckDuckGo, X/Twitter, Reddit, GitHub, YouTube, ArXiv, Wikipedia, and more)
News-oriented search tuned to recent articles
On-demand crawling: fetch full page content for selected results
Sitemap / link discovery for a given site
Reasoning tool endpoint (DeepSeek R1) for longer-form problem solving
Trending endpoints for GitHub and Hacker News
Remote (HTTP) and local (stdio via npx) deployment options
Simple API key authentication: Authorization header or query param
Easy integration with MCP-enabled clients (Claude Desktop, VS Code extensions, Cursor, etc.)

Installation / Configuration

Remote (no local install)

Use the hosted MCP URL and send your API key either as an Authorization header or as a query parameter.

Example: Authorization header (preferred)

{
  "mcpServers": {
    "search1api": {
      "url": "https://mcp.search1api.com/mcp",
      "headers": {
        "Authorization": "Bearer YOUR_SEARCH1API_KEY"
      }
    }
  }
}

Example: Query parameter authentication

https://mcp.search1api.com/mcp?apiKey=YOUR_SEARCH1API_KEY

Client-specific examples

Claude Desktop / Cursor / VS Code MCP config (JSON snippets above apply).
Claude.ai (web): add a custom connector using the URL including ?apiKey=…
Windsurf: set serverUrl to “https://mcp.search1api.com/mcp?apiKey=YOUR_SEARCH1API_KEY”
Claude Code (CLI): add an HTTP MCP transport with an Authorization header:

claude mcp add --transport http search1api https://mcp.search1api.com/mcp \
  --header "Authorization: Bearer YOUR_SEARCH1API_KEY"

Local (stdio) mode

Run the MCP server locally via npx; useful if you want to keep traffic on-device or avoid remote calls.

{
  "mcpServers": {
    "search1api": {
      "command": "npx",
      "args": ["-y", "search1api-mcp"],
      "env": {
        "SEARCH1API_KEY": "YOUR_SEARCH1API_KEY"
      }
    }
  }
}

Agent / CLI skill

Agent-style skill and CLI helpers are available through the search1api-cli package:

npm install -g search1api-cli
npx skills add fatwang2/search1api-cli

Available Tools

The MCP server exposes a set of tools (actions) you can invoke from your agent. Core tools are described below.

Purpose: general web search across multiple providers
Key parameters:

Parameter	Required	Default	Notes
query	Yes	—	Search query text
max_results	No	10	Number of results to return
search_service	No	google	Options: google, bing, duckduckgo, yahoo, x, reddit, github, youtube, arxiv, wechat, bilibili, imdb, wikipedia
crawl_results	No	0	Number of top results to fetch full HTML/text
include_sites	No	[]	Whitelist domains
exclude_sites	No	[]	Blacklist domains
time_range	No	—	day, month, year

news

Purpose: news-focused article search
Key parameters: query (required), max_results (default 10), search_service (default bing; options include google, bing, duckduckgo, yahoo, hackernews), crawl_results, include/exclude sites, time_range

crawl

Purpose: fetch and extract content from a specific URL
Parameters: url (required)

sitemap

Purpose: enumerate links related to a URL / site
Parameters: url (required)

reasoning

Purpose: run longer-form reasoning using DeepSeek R1
Parameters: content (required) — the question or task to process

trending

Purpose: fetch trending items
Parameters: search_service (github, hackernews), max_results (default 10)

Use Cases

Enrich an LLM answer with live web citations

Flow: Agent issues a search tool call with query=“latest GPU releases 2026”, max_results=5, crawl_results=2. The MCP server returns structured search hits plus full text for the top 2 pages to extract quotes and cite URLs.

News monitoring for a topic

Example: call the news tool with query=“banking fraud detection”, time_range=day, max_results=20 to surface recent articles from Bing/Hacker News. Optionally set include_sites to preferred outlets.

Extract article text for ingestion

Example: use the crawl tool to fetch a URL and return the cleaned article body, metadata (title, author, published date), and main images for downstream indexing.

Site map and link discovery for scraping

Example: call sitemap with url=“https://example.com” to retrieve discovered links, filter by path prefix, and schedule deeper crawls for selected pages.

Example JSON payload for a search tool call (for MCP-capable clients):

{
  "tool": "search",
  "params": {
    "query": "quantum computing news",
    "max_results": 8,
    "search_service": "google",
    "crawl_results": 2,
    "time_range": "month"
  }
}

Notes and Licensing

Authentication: prefer the Authorization: Bearer header for per-request scoping; query parameter is supported for convenience.
The repository and server are MIT licensed. For agent integration, see the separate search1api-cli package for skill tooling and examples.

If you are new to MCP-based connectors, start with the hosted URL + Authorization header configuration in your MCP client, test simple search calls, then enable crawl_results incrementally to fetch page content.

Search1API MCP Server — Web Search, News, Crawling