SerpKite

io.github.SerpKitev1.0.0更新于 Oct 5, 2026

Google search, news, maps, scholar and public webpages for AI agents, with Markdown results.

已验证Streamable HTTP可网页运行Web Search & ScrapingKnowledge & Memory

概览

AI 生成的概览

让助手执行 Google 网页、新闻、地图、学术、专利、购物、图片和视频搜索，并把公开页面转换为 Markdown。

功能
SerpKite 是一个托管的远程 MCP 服务器，提供十四个工具：search、news、maps、scholar、patents、shopping、images、videos、autocomplete、webpage、extract、map、crawl 和 crawl_result。搜索类工具返回 Google 结果、答案、知识图谱和相关问题；webpage 和 extract 把公开的 HTML 或 PDF 转成 Markdown，extract 还能按查询筛选段落。爬取类工具可启动对公开站点页面的爬取并读取状态与结果。结果以适合大模型的 Markdown 返回。
适用场景
当助手需要实时的 Google 搜索结果、新闻、地点信息、学术论文、专利、购物或媒体结果，或需要读取并提取公开网页与站点内容时使用。它是自建搜索或抓取方案之外的托管选择。
运行要求
需要可访问网络的远程 Streamable HTTP 端点 SerpKite API 密钥，并在 Authorization 请求头中以 Bearer 加密钥的形式发送。客户端必须支持自定义请求头，否则需要基于 Node.js 的 stdio 桥接。无需安装 npm 包。
安装前请注意
Authorization 请求头携带密钥，请勿提交到源码仓库，优先使用环境变量或密码输入。按额度计费，价格与对应的 REST 端点一致，搜索可能产生费用；可为专用密钥设置每月额度上限以控制代理支出。crawl 和 extract 会抓取第三方公开页面，请求会发送到 SerpKite 服务。

安装

在 SourceWeft 中

  1. 打开 控制台中的 SerpKite，将其添加到工作区。
  2. 为需要使用其工具的对话启用该服务。

Web executable，通过 Streamable HTTP。 远程服务在工作区中配置后即可从网页运行时运行。

其他 MCP 客户端

把它添加到你客户端的 mcpServers 配置中。

{
  "mcpServers": {
    "serpkite": {
      "type": "http",
      "url": "https://api.serpkite.com/v1/mcp"
    }
  }
}

README

SerpKite MCP

Give AI agents Google search and public web content through the SerpKite remote Model Context Protocol server. Maintained by the SerpKite team.

This repository contains connection examples and registry metadata for the hosted service. The API implementation is proprietary; the documentation and configuration in this repository are MIT licensed.

Connection

  • Endpoint: https://api.serpkite.com/v1/mcp
  • Transport: stateless Streamable HTTP
  • Authentication: Authorization: Bearer <your-api-key>
  • Protocols: 2025-06-18, 2025-03-26, 2024-11-05

Get an API key at app.serpkite.com. Read the pricing and MCP documentation. Credits never expire; failed and empty searches are free. Set a dedicated key's monthly credit limit to control agent spending.

OAuth is planned. Clients must currently support custom headers or use the stdio bridge below. No SerpKite npm MCP package is required.

Claude Code

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

Cursor

Add to ~/.cursor/mcp.json or your project's .cursor/mcp.json:

json
{  "mcpServers": {    "serpkite": {      "url": "https://api.serpkite.com/v1/mcp",      "headers": { "Authorization": "Bearer ${env:SERPKITE_API_KEY}" }    }  }}

VS Code

Add to .vscode/mcp.json. The password input keeps the API key out of the committed file:

json
{  "inputs": [    { "type": "promptString", "id": "serpkite-key", "description": "SerpKite API key", "password": true }  ],  "servers": {    "serpkite": {      "type": "http",      "url": "https://api.serpkite.com/v1/mcp",      "headers": { "Authorization": "Bearer ${input:serpkite-key}" }    }  }}

Cline

Follow llms-install.md for Cline's remote connection configuration.

Claude Desktop and stdio clients

With Node.js installed, use the third-party mcp-remote bridge in your client's MCP configuration:

json
{  "mcpServers": {    "serpkite": {      "command": "npx",      "args": ["-y", "mcp-remote", "https://api.serpkite.com/v1/mcp", "--header", "Authorization: Bearer ${SERPKITE_API_KEY}"],      "env": { "SERPKITE_API_KEY": "YOUR_API_KEY" }    }  }}

Keep API keys out of source control. ChatGPT connectors that only support OAuth or no authentication cannot currently connect directly.

Tools

ToolPurpose
searchGoogle web results, answers, knowledge graph and related questions
newsGoogle News articles
mapsPlaces, ratings, addresses and coordinates
scholarAcademic papers, citations and PDF links
patentsPatent search
shoppingProducts, prices and merchants
imagesImage search
videosVideo search
autocompleteQuery suggestions
webpagePublic HTML or PDF to Markdown
extractRead multiple URLs, optionally select passages by query
mapDiscover a site's URLs
crawlStart a crawl of public site pages
crawl_resultRead crawl status and results

Tool results are LLM-ready Markdown. Pricing follows the matching REST endpoints. Only public, logged-out pages are fetched. See tool inputs and costs.

Verify the connection

These metadata requests do not run searches:

bash
curl https://api.serpkite.com/v1/mcp \  -H "Authorization: Bearer $SERPKITE_API_KEY" \  -H "Content-Type: application/json" \  -d '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2025-06-18","capabilities":{},"clientInfo":{"name":"curl","version":"1.0"}}}'
curl https://api.serpkite.com/v1/mcp \  -H "Authorization: Bearer $SERPKITE_API_KEY" \  -H "Content-Type: application/json" \  -d '{"jsonrpc":"2.0","id":2,"method":"tools/list"}'

The remote endpoint and all fourteen tool names were verified on 2026-10-05.

Support

Documentation · Status · [email protected]

Google is a trademark of Google LLC. SerpKite is not affiliated with Google.

来源：README.md，提交 ee407da

工具

0
工具元数据尚未被收录。

版本历史

1
  1. v1.0.0最新Oct 5, 2026