Search1api MCP

What fatwang2/search1api-mcp MCP server does

fatwang2/search1api-mcp MCP server exposes Search1API capabilities as MCP tools for researching public online content. The service can search the web or news, retrieve readable content from a supplied page, enumerate links associated with a URL, and report currently popular repositories or stories from GitHub and Hacker News.

Search results include source-oriented fields such as an identifier, title, URL, and text or snippet. An agent can pass a returned URL to crawl when a result summary is not enough. The tools are intended for requests that depend on current information, recent coverage, public documentation, or discovery of pages within a site.

How it works

The hosted endpoint is available at https://mcp.search1api.com/mcp over Streamable HTTP. OAuth-aware clients can connect and complete browser sign-in, while other clients can send a Search1API key as a Bearer token. A legacy query-string key is also supported, though the documentation recommends OAuth or an Authorization header because URL credentials can appear in logs and shell history.

For local use, an MCP client starts the package with npx -y search1api-mcp. Node.js 20 or newer is required. The local process reads SEARCH1API_KEY from its environment when tools are called. Self-hosted HTTP deployments can restrict accepted hostnames and browser origins with MCP_ALLOWED_HOSTS and MCP_ALLOWED_ORIGINS.

Every MCP request requires credentials in the remote setup, including initialization and tool discovery. Unauthenticated requests receive an OAuth challenge, allowing compatible clients to begin sign-in. The server card remains available for pre-connection inspection.

Setup and configuration

For a remote connection, configure the endpoint URL in the MCP client and choose OAuth when supported. API-key configuration uses the following header pattern:

Authorization: Bearer YOUR_SEARCH1API_KEY

A local stdio configuration can use this structure:

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

The README provides client-specific examples for Claude Desktop, Cursor, VS Code, Claude Code, and Windsurf. Cursor can use the repository's plugin files for a remote OAuth connection, or a manual MCP configuration. For local Cursor testing, plugin files must be placed in the expected local plugin directory rather than linked through an external symlink.

Tools and capabilities

• search queries the web. It supports result limits, selectable services including Google, Bing, DuckDuckGo, Yahoo, X, Reddit, GitHub, YouTube, arXiv, WeChat, Bilibili, IMDb, and Wikipedia, site filters, time ranges, and optional crawling of top results.
• news searches news sources with Google, Bing, DuckDuckGo, Yahoo, or Hacker News. It supports site filters, time ranges, result limits, and optional extraction of leading articles.
• crawl extracts a readable title and full text from a public URL.
• sitemap returns related public links for a URL or domain exploration task.
• trending lists popular repositories or stories from GitHub or Hacker News.

Search and news requests have a base one-credit cost, with each successfully crawled result adding one credit according to the documented pricing behavior.

Limitations and notes

fatwang2/search1api-mcp MCP server requires access to Search1API and, for API-key usage, a paid API key from its dashboard. Query-parameter authentication is available for compatibility but is less safe than OAuth or a request header.

The available search services and filters differ by tool, so callers should select the appropriate tool rather than expect every backend in both web and news search. Trending data is limited to GitHub and Hacker News. Local deployments also need Node.js 20 or newer, while self-hosted browser-facing HTTP deployments may need explicit allowed-host and origin settings.