Skip to content

New AI assistant answers: ChatGPT, Gemini, Copilot and AI Mode by API

SerpKite
Get API key

Remote MCP · streamable HTTP

Google Search MCP server

Give Claude, Cursor, VS Code, Windsurf or your own agent live Google search with one line of config. The SerpKite MCP server runs at https://api.serpkite.com/v1/mcp: nothing to install, authenticate with your API key.

Overview

What the SerpKite MCP server is

The Model Context Protocol (MCP) is the standard way for AI apps to discover and call tools. SerpKite hosts a remote MCP server over streamable HTTP, so any client that supports remote servers can use Google search without you running a process locally.

Authenticate with the header Authorization: Bearer skt_live_…. Each tool call is billed like the matching API endpoint, and failed calls are free.

Setup

Set it up in 3 steps

  1. 1

    Get an API key

    Sign up (no card), create a key in the dashboard and export it as SERPKITE_API_KEY. New accounts get 2,500 free credits, then 1,000 every month.

    export SERPKITE_API_KEY=skt_live_…
  2. 2

    Add the server to your client

    Pick your client below. Claude Code is one command; Cursor, VS Code and Windsurf take a few lines of JSON.

    claude mcp add --transport http serpkite https://api.serpkite.com/v1/mcp \
      --header "Authorization: Bearer $SERPKITE_API_KEY"
  3. 3

    Ask a question that needs the web

    Try "What changed in the latest Python release? Search the web and cite sources." The client lists SerpKite's tools and calls search on its own.

Code

Client configuration

Replace skt_live_… with your key, or reference an environment variable where the client supports it.

# Add for the current project (use --scope user to add it everywhere)
claude mcp add --transport http serpkite https://api.serpkite.com/v1/mcp \
  --header "Authorization: Bearer $SERPKITE_API_KEY"

# Check it's connected
claude mcp list
  • Cursor: ~/.cursor/mcp.json
  • VS Code: .vscode/mcp.json
  • Windsurf: ~/.codeium/windsurf/mcp_config.json
  • Claude Desktop: claude_desktop_config.json

Tools

What the server exposes

  • search

    Google web results as Markdown (1 credit).

  • news

    Google News articles with source and date (1 credit).

  • maps

    businesses from Google Maps with rating, hours, phone and website (1 credit).

  • scholar

    papers with citation counts and PDF links (1 credit).

  • webpage

    any public URL as clean Markdown (1 credit).

  • answer

    a short answer from live Google results with numbered citations, written by Claude (3 credits fast, 8 standard; free if the model fails).

  • research and research_result: an async deep-research task that ends in a cited report (30, 75 or 200 credits, charged only when it's written).

  • chatgpt, gemini, copilot, ai_mode and brave_ai: ask an AI assistant as a logged-out user and get its answer with cited sources, products and mentions of the brands you track (2–4 credits per answer; free if it fails).

  • extract, map, crawl and crawl_result: URL lists, site maps and whole sites as Markdown, with optional summaries and JSON for your schema. 23 tools in all; tools/list shows every input.

Claude Desktop

Claude and Claude Desktop: add a custom connector

In Claude (web or desktop), open Settings → Connectors → Add custom connector and enter https://api.serpkite.com/v1/mcp. Claude opens a SerpKite sign-in page; approve the connection and the tools appear. Calls are billed to your credits under a key named "Claude (OAuth)", which you can limit or revoke on the Keys page.

Claude Desktop's JSON config launches local (stdio) servers. To reach a remote server that authenticates with a header, run the open-source mcp-remote bridge with npx, as in the Claude Desktop tab above. It needs Node.js 18 or newer.

The Authorization:$SERPKITE_AUTH pattern (no space after the colon, value taken from env) avoids an argument-quoting issue on some platforms. Restart Claude Desktop after editing the file.

ChatGPT and the OpenAI API

ChatGPT connectors and the Responses API

ChatGPT's custom connectors (Developer mode) connect to remote MCP servers that use OAuth. Add https://api.serpkite.com/v1/mcp as the connector URL with OAuth authentication, then sign in to SerpKite and approve the connection; no key is pasted anywhere.

If you build on the OpenAI API, the Responses API's mcp tool lets you pass the Authorization header yourself, so your own app can use an API key:

OpenAI Responses API + SerpKite MCP
import os
from openai import OpenAI

client = OpenAI()
resp = client.responses.create(
    model="gpt-5-mini",
    tools=[{
        "type": "mcp",
        "server_label": "serpkite",
        "server_url": "https://api.serpkite.com/v1/mcp",
        "headers": {"Authorization": f"Bearer {os.environ['SERPKITE_API_KEY']}"},
        "require_approval": "never",
    }],
    input="What are this week's biggest AI funding rounds? Cite sources.",
)
print(resp.output_text)

FAQ

MCP server and SerpKite: common questions

Start building

Connect your AI client to Google

2,500 free credits, then 1,000 every month. No credit card.