Hashscraper

io.github.hashscraper由 GitHub 账号 hashscraper 发布v0.1.0更新于 Oct 29, 2026

Coupang top 30 and public web data bots (Naver, Instagram, Olive Young, maps, reviews) for AI agents

概览

AI 生成的概览

Hashscraper 让 AI 助手无需编写爬虫代码即可采集公开网页数据,包括 Coupang 前 30 榜单和 120 多个现成数据机器人。

可以做什么
  • 连接与账户:通过浏览器登录流程连接账户,查询账户邮箱和当日剩余免费额度
  • 浏览机器人目录:搜索 120 多个内置数据机器人及其输入参数
  • 执行采集:运行机器人并返回数据行,或按数据集 ID 取回此前运行的结果
  • Coupang 榜单:列出提供每日前 30 的类目,并获取某类目前 30 的价格、评论、配送和详情页字段
  • 导出结果:将较大结果保存为本地 CSV 文件
适用场景
适合让助手采集公开的电商、社交、地图、评论或社区数据,尤其是 Coupang 畅销榜以及 Naver、Instagram、Olive Young、Musinsa、Kurly 等韩国平台的数据,并整理成摘要或 CSV,而无需自行开发爬虫。
运行要求
通过 npx 以 stdio 在本地运行,需要 Node.js 20 及以上版本。需要 HashScraper 账户:可在浏览器中通过 login 工具一次性连接,或设置可选的 HASHSCRAPER_API_KEY 环境变量。需要能访问 HashScraper API 的网络。
安装前请注意
采集会消耗额度:每日免费额度在韩国时间午夜重置且不累积,付费额度仅在明确要求时使用。首次使用额度的命令前服务器会暂停,需你接受数据条款,条款限定数据仅供内部分析,不得转售或再分发。API 密钥保存在本地凭据文件中,并与 HashScraper Agent Skill 共用。数据会发送至 HashScraper API。

安装

在 SourceWeft 中

  1. 打开 控制台中的 Hashscraper,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。

Desktop only,通过 STDIO。 STDIO 服务会启动本地进程,因此需要 SourceWeft 桌面宿主。

其他 MCP 客户端

参照 仓库 中的启动说明。

README

HashScraper MCP

Collect public web data from your AI agent. Ask "Summarize today's prices of the Coupang top 30 for 커피믹스" and the HashScraper MCP server fetches the daily best-seller ranking of the category — with price, reviews, delivery and detail-page columns — or runs one of 120+ ready-made data bots (Naver blog/cafe/news/shopping, Instagram, YouTube, Olive Young, Musinsa, Kurly, KREAM, Google/Kakao maps, app stores, community boards …).

  • No scraper code and no browser on your side. Node.js 20+ and one line of config.
  • Every account gets free credits every day (refill at midnight KST). The server uses only the daily free credits unless you tell it to use your own.
  • Your API key is saved on your computer only (~/.config/hashscraper/credentials, readable only by you) and is never shown to the agent.

Homepage: https://www.hashscraper.com/mcp?utm_source=mcp&utm_medium=github&utm_campaign=mcp_repo

Install

The server runs locally over stdio: npx -y @hashscraper/mcp.

Claude Code

bash
claude mcp add hashscraper -- npx -y @hashscraper/mcp

Claude Desktop

Add to claude_desktop_config.json, then restart Claude Desktop:

json
{  "mcpServers": {    "hashscraper": {      "command": "npx",      "args": ["-y", "@hashscraper/mcp"]    }  }}

Cursor

Add to ~/.cursor/mcp.json (or the project's .cursor/mcp.json):

json
{  "mcpServers": {    "hashscraper": {      "command": "npx",      "args": ["-y", "@hashscraper/mcp"]    }  }}

Codex

Add to ~/.codex/config.toml:

toml
[mcp_servers.hashscraper]command = "npx"args = ["-y", "@hashscraper/mcp"]

MCP Bundle (.mcpb)

To build a one-file bundle for apps that install .mcpb files (it includes the server and its dependencies, and asks for the optional API key):

bash
npm install && npm run pack:mcpb   # -> server.mcpb

Connect once

Ask your agent to "connect HashScraper" (or just ask for some data). It calls the login tool, which returns a link and an 8-character code. Open the link, sign in (or sign up for free with Google or email), check that the code matches, and click Approve and connect. Then tell the agent you approved — it calls login again and the key is saved. You never copy or paste the key.

Prefer your own key? Create one in My page → API and set HASHSCRAPER_API_KEY in the server's environment. It is used before the saved key.

Try these prompts

  1. "Summarize today's prices of the Coupang top 30 for 커피믹스."
  2. "Which Coupang categories can you rank? Compare the top 5 of 에센스 by price."
  3. "Collect 50 Naver blog posts about 성수동 카페 and save them as a CSV."
  4. "What is the price of this Coupang product right now? https://www.coupang.com/vp/products/…"
  5. "How many free credits do I have left today?"

Tools

ToolWhat it doesCredits
loginConnect your HashScraper account in the browser (two calls: link, then finish)free
meAccount email and today's free creditsfree
catalogSearch the built-in list of 120+ bots and their inputsfree
runCollect with one bot and return the rows (waits up to 55 s, then gives a dataset id)per row, see catalog
fetchGet the rows of an earlier run by dataset idfree
coupang_categoriesCoupang categories with a ready daily top 30free
coupang_topTop 30 of a Coupang category with detail columns (collected once a day)300 per category per snapshot; fetching the same snapshot again is free

Large results can be saved to a file with csv_path (an absolute path ending in .csv, UTF-8 with BOM so Excel reads Korean).

Free credits

  • Free credits refill every day at midnight (KST). Unused credits do not carry over.
  • When today's free credits run out, the tool stops and tells you when they refill. Your own credits are used only when you ask (use_credits).

Data terms

The data is for your internal analysis only — no resale or redistribution: https://www.hashscraper.com/data-terms

Before the first command that uses credits, the server stops and shows the terms. Your agent must show them to you and may continue only after you clearly agree (accept_terms). It is asked once per account. The agent never agrees for you.

Privacy

The server collects public pages only and never asks for a site login. It talks only to the HashScraper API (https://www.hashscraper.com/v5/api). The saved key lives in ~/.config/hashscraper/credentials (owner-only) and is shared with the HashScraper Agent Skill. Delete the file to disconnect, or turn the key off in My page → API & integrations.

Need something else?

Missing site or Coupang category, daily scheduled collection, or large volumes? Email [email protected] with the site, the data you need, how much and how often — or talk to us: https://www.hashscraper.com/quotes/new?utm_source=mcp&utm_medium=github&utm_campaign=mcp_repo

License

MIT. See LICENSE.

来源:README.md,提交 1633536

工具

0
SourceWeft 尚未在沙箱中启动过该软件包,因此其工具未知。

版本历史

1
  1. v0.1.0最新Oct 29, 2026

评分

已满足适用检查项的 50%

该评分衡量此收录条目的完整程度,依据注册中心条目、端点与公开来源计算得出。它不是对服务器本身的评价,也不表示服务器安全或当前可用:可连通性是带日期的检查结果,评分也不包含安装量或用户评价。

  • 近 30 天内可连通25 分

    请保持端点能够响应 MCP 握手;SourceWeft 每 30 天探测一次。

  • 已声明认证方式10 分

    请在注册中心条目中声明凭据请求头,或保持端点能够响应探测,以便判定其认证方式。

  • 已列出工具15 分

    请让服务器在无凭据时响应 tools/list,或发布可在沙箱中无凭据启动的软件包。

  • 描述必需15 分
  • 图标10 分
  • 文档15 分
  • 许可证10 分