Add Web Search to Any MCP Client

The Problem

The Model Context Protocol lets AI assistants in Claude Desktop, Cursor, Windsurf, and other MCP-compatible clients call external tools. But getting web search into an MCP client typically means running a local server, managing browser automation, dealing with API key rotation for multiple providers, and writing glue code that translates between the MCP tool schema and whatever search backend you chose. The setup is brittle, undocumented for most providers, and breaks when the MCP spec evolves. Most developers give up and use their AI assistant without web access, which means every answer is limited to the model's training cutoff.

The Scavio Solution

Scavio ships an official MCP server that you can add to Claude Desktop, Cursor, or Windsurf in minutes. The server exposes web search, product search, video search, and Reddit search as MCP tools with typed schemas. Your AI assistant can search Google, Amazon, YouTube, Walmart, and Reddit mid-conversation and ground its answers in live data. The setup is a single entry in your MCP configuration file pointing to the Scavio server, with your API key as the only required parameter. No browser automation, no local proxy, no glue code.

Who It Is For

Developers using Claude Desktop, Cursor, or Windsurf who want their AI assistant to search the live web. If you have tried to set up web search in an MCP client and gave up because the setup was too complex, this is the three-minute path.

Key Benefits

• Official MCP server with typed tool schemas
• Works with Claude Desktop, Cursor, and Windsurf
• Five platforms available as MCP tools from one server
• Three-line configuration, no local infrastructure
• Grounded answers with live web data inside the IDE

Python Example

# mcp_config.json - Add to your MCP client configuration
# {
#   "mcpServers": {
#     "scavio": {
#       "command": "npx",
#       "args": ["-y", "@anthropic-ai/scavio-mcp-server"],
#       "env": { "SCAVIO_API_KEY": "your_scavio_api_key" }
#     }
#   }
# }

# Once configured, use Scavio in any Python agent with MCP support
import requests

API_KEY = "your_scavio_api_key"

def mcp_search(query: str, platform: str = "google"):
    r = requests.post(
        "https://api.scavio.dev/api/v1/search",
        headers={"x-api-key": API_KEY},
        json={"platform": platform, "query": query},
        timeout=10,
    )
    return r.json().get("organic", [])[:5]

print(mcp_search("MCP server setup guide 2026"))

JavaScript Example

// mcp_config.json - Add to your MCP client configuration
// {
//   "mcpServers": {
//     "scavio": {
//       "command": "npx",
//       "args": ["-y", "@anthropic-ai/scavio-mcp-server"],
//       "env": { "SCAVIO_API_KEY": "your_scavio_api_key" }
//     }
//   }
// }

const API_KEY = "your_scavio_api_key";

async function mcpSearch(query, platform = "google") {
  const r = await fetch("https://api.scavio.dev/api/v1/search", {
    method: "POST",
    headers: {
      "x-api-key": API_KEY,
      "content-type": "application/json",
    },
    body: JSON.stringify({ platform, query }),
  });
  const data = await r.json();
  return (data.organic ?? []).slice(0, 5);
}

console.log(await mcpSearch("MCP server setup guide 2026"));

Platforms Used