Open-source B2B and marketplace ecommerce; new stores include an MCP server for AI agents.
KlickyMart
概览
作者提供的说明(英文原文)
Which online stores AI shopping agents can buy from: guest checkout, pay methods, landed cost.
PopNx MCP 的三大支柱
这三项永远不会合并成一个分数:意见、测量和人气是不同的东西。
- 用户评价i暂无评价评价功能将在后续阶段与账户功能一起开放
- 我们的基础检查i3 项中通过 3 项依据公开信息检查;尚未扫描代码
- 使用情况i尚未统计★ –
它能帮你做什么
我们还没有为它写通俗的介绍。下面是我们从注册中心读取到的事实。
- 连接方式: 在云端运行
- 传输方式: Streamable HTTP
- 分类: 网店与支付
安装
选择你使用的 AI 应用,然后复制设置。
安装之前
- 基础检查:3 项中通过 3 项
- 在作者的服务器上运行:你向它提出的内容会经过他们的服务器。
- 可以做 15 件事(工具) search_merchants, get_record, check_domain, find_offers …
- 注册中心没有列出需要填写的密钥。
你需要
- 在云端运行无需安装任何东西,通过互联网连接
你用的是哪个应用?
在上方选择你的应用。我们会记住它,下次按钮会显示“安装到 …”。
适用于网页版、桌面版和手机版 Claude:
- 打开 Settings → Connectors
- 点击 Add custom connector
- 粘贴下面的地址,然后点击 Add
- 如果要求登录,请登录作者的服务
https://klickymart.com/mcp在终端中运行这条命令
claude mcp add --transport http klickymart https://klickymart.com/mcp或者自己复制设置
粘贴到项目中的 .cursor/mcp.json,或 ~/.cursor/mcp.json
{
"mcpServers": {
"klickymart": {
"url": "https://klickymart.com/mcp"
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
或者自己复制设置
粘贴到项目中的 .vscode/mcp.json
{
"servers": {
"klickymart": {
"type": "http",
"url": "https://klickymart.com/mcp"
}
}
}浏览器会先询问是否打开应用,然后应用会显示设置并请你确认。密钥的值会留空,由你自己填写。
ChatGPT 只能连接互联网上的服务器:
- 打开网页版 ChatGPT → Settings → Connectors (Apps)
- 点击 Create 或 Add(部分套餐需要先在 Advanced 中开启 Developer mode)
- 粘贴下面的地址,起个名字,然后保存
- 如果要求登录,请登录作者的服务
https://klickymart.com/mcp菜单名称可能因套餐和版本不同而略有差异。
你的选择通过 Cookie 保存在这台设备上,无需账户。
这些设置是根据作者在注册中心的条目生成的。如果 README 的写法不同,请以 README 为准。
它能做什么
工具名称和描述与服务器在 2026年11月6日 报告的完全一致。这些内容由作者撰写,不是我们写的;我们以纯文本形式显示。
search_merchantsSearch KlickyMart merchant records (same filters as GET /q)
get_recordFetch one merchant record by domain
check_domainCan my agent shop here? Verdict for one domain from the index (reachable, checkout-blocked, agents-not-welcome, blocked-tos, error, not-in-index; see verdict_reason and primary_reason), with checkout depth, pay methods, BTC grade, record time and evidence URL. Never triggers a li
find_offersProduct-level offers KlickyBot observed on store pages (schema.org JSON-LD or Shopify JSON): price, currency, stock, URL, checked_at. Match by gtin (UPC/EAN), mpn, or name words (q). Optional zip adds a landed-cost estimate; live=true also returns a live (uncached) Shopify Catalo
get_landed_priceOne call: a product's landed price at a US ZIP. Searches KlickyMart-observed offers (and, with live=true, the live Shopify Catalog, never stored), and returns for every product: price, stock, url, checked_at, shipping (with source), est. tax with the rate used, est_total ONLY whe
build_experience_noteReturns an UNSIGNED Nostr kind 422 'shopping experience' event template for a store, for you to sign with your own (ideally site-specific) key and publish to your own relays. KlickyMart never signs or publishes for you. Optionally POST the signed event to /reviews (or the submit_
get_checkout_mapCheckout steps our probe reached for a store (stopping at the payment screen, never submitting), the first blocker (bot-wall, http-refusal, network-error, no-product-link, no-add-to-cart, form-gate, login-wall, robots, unclassified, none) and landed-cost signals.
recheck_domainRe-read robots.txt and the home-page status for an indexed store now. Max 1 per store per 10 minutes (else cached result + next_allowed_at), 60 per hour overall.
watch_itemDRAFT (GO D8, not served yet): returns an unsigned kind 420 watch template for price/stock changes. KlickyMart does not act on watches yet and never signs for you.
find_alternativesWhere else can my agent buy this? Give a blocking domain and/or a category (electronics, gift-cards, groceries, apparel, hardware, books, travel, crypto-hardware, other, or a product word like headphones). Product words resolve to a finer sub (headphones -> audio). Returns in-map
find_agent_nativeAgent-native stores: sellers an agent can pay without a browser checkout (L402/x402 paid APIs, VPN, eSIM, VPS, gift cards, nostr-native, agent-commerce). Each item has agent_native {category, label verified-402|claimed, rails, rails_status, probe_402 with price_seen}. verified-40
get_reviewsKind 422 reviews of a merchant (signed claims by pseudonymous keys; not verified by KlickyMart).
submit_reviewSubmit a kind 422 shopping-experience note or review you already signed locally (never send a secret key). Needs a store tag (a=31420: : or r= ) and depth (browse|product|cart|checkout|shipping|payment|paid|delivered; legacy attempt accepted) or rating. Optional: basis (visit|dir
submit_storeSuggest a store KlickyMart has not indexed yet. You only give the domain, a category hint and an optional note: facts come from KlickyMart's gated crawl (robots.txt and ToS first), never from you. Only the bare registrable domain is kept (paths, queries, affiliate/referral parame
wallet_setup_guideNon-custodial wallet setup guide for agents: Nostr key, self-custodial or agent-friendly Lightning options (NWC, Alby Hub, phoenixd, LNbits) with who holds the funds, how to get a lud16 for tips, Cashu wallets and mint risk, plus a checklist. KlickyMart never generates, holds or
基础检查i
我们检查的是任何人都能看到的信息;我们没有阅读代码。不适用的检查项不计入总数。 · 检查于 2026年11月6日
- ✓ 通过
作者与仓库一致
注册名称已通过域名 klickymart.com 验证(注册中心会核实域名所有权)
- … 尚未检查
仍在维护
我们还没有从 GitHub 读取这个仓库
- … 尚未检查
有许可证
我们还没有从 GitHub 读取这个仓库
- ✓ 通过
端点有响应
在 626 毫秒内响应(HTTP 200)
- – 不适用
使用 OAuth 登录
无需登录
- ✓ 通过
工具描述中没有隐藏指令
已读取 15 个工具描述,未发现针对 AI 的隐藏指令
与“网店与支付”中的其他服务器比较i
- 星标数多于此分类中其他 1745 个服务器的 0%
- 发布到注册中心的时间比此分类中其他 1745 个服务器的 80% 更新
- 基础检查结果等于或优于此分类中其他 1745 个服务器的 100%
与注册中心里这个分类中所有活跃的服务器比较。只有当发布者拥有该仓库时,星标数才计入。
常见问题
KlickyMart 安全吗?
它在 3 项适用的基础检查中通过了 3 项。我们没有阅读代码。在添加密钥或令牌之前,请先看看这个服务器要求哪些权限,并使用只具备所需权限的密钥。
它需要哪些密钥或代码?
注册中心没有列出必需的密钥,不过有些服务器会在首次使用时要求你登录。
它在我的电脑上运行还是在云端?
它在作者的云端运行。通过互联网连接,无需安装,但你发送的内容会经过作者的服务器。
它还在维护吗?
GitHub 仓库最后更新于–,最新版本(2026.10.31)于6天前进入注册中心。
哪些应用可以使用它?
支持 MCP 的应用,例如 Claude Desktop、Claude Code、Cursor 和 VS Code。它通过 URL 连接,所以接受远程 MCP 服务器的应用也可以使用,例如开启开发者模式的 ChatGPT。
有什么可以替代它?
同一分类中有:Spree Commerce: open-source B2B and marketplace ecommerce, Stripe MCP, MCP Cn Commerce
替代选择
同一分类中其他作者的服务器,按星标数排序。
如何看卡片:★ GitHub 星标i三大支柱:评价 · 检查 · 使用情况i更新 = 最新版本发布到注册中心的时间i
Stripe MCP
MCP server integrating with Stripe - tools for customers, products, payments, and more.
MCP Cn Commerce
8 read-only MCP platform entries; support is operation-specific; merchant live acceptance pending
作者的 README
打开 README
按原文从作者的仓库复制。PopNx MCP 没有审查或背书其内容。HTML、图片和脚本已被移除。链接会离开本站。
KlickyMart
KlickyMart is a public, signed map of online stores for software agents: whether an agent can reach a store, how checkout works, and which payment methods the store accepts, with attention to Bitcoin and privacy-respecting Bitcoin options. Records are published as Nostr events (see NIP-DRAFT.md). A FloppyKit project.
Status: draft. No crawl data is published in this repository yet.
Contents
NIP-DRAFT.md: draft event kinds for the directory and payment handshakes.CRAWL-RUBRIC.md: record schema and grading rules (v0.3 draft changes at the end).CASCADE.md: build plan.crawler/: Playwright crawler (crawl.py), deterministic payment extractor (extract.py), terms-of-service checker (tos.py), signing helpers, tests.crawler/KNOWN-LIMITATIONS.md: known limitations.site/bot/index.html: the crawler info and opt-out page, served at https://klickymart.com/bot.site/: KlickyMart web app (Worker, API, dashboard). See site/README.md.
Use it from an agent (MCP)
Agents should start at https://klickymart.com/llms.txt. MCP endpoint: https://klickymart.com/mcp. It uses a minimal, stateless Streamable HTTP transport: one JSON-RPC message per POST, answered with a single application/json body. It negotiates protocol versions 2024-11-05, 2025-03-26 or 2025-06-18 (tools only). There is no SSE stream and no session ID, and GET /mcp returns 405. Tools: search_merchants, get_record, check_domain. No key needed.
Claude Desktop (claude_desktop_config.json launches stdio servers, so this goes through the mcp-remote bridge, which needs Node.js):
{
"mcpServers": { "klickymart": { "command": "npx", "args": [ "-y", "mcp-remote", "https://klickymart.com/mcp" ] } } }
Cursor (~/.cursor/mcp.json or .cursor/mcp.json):
{
"mcpServers": { "klickymart": { "url": "https://klickymart.com/mcp" } } }
Any Streamable HTTP client:
curl -s https://klickymart.com/mcp -H "Content-Type: application/json" -H "Accept: application/json, text/event-stream" \\
-d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"check_domain","arguments":{"domain":"example.com"}}}'
Reviews: anyone can review a merchant with a signed kind 422 event (a tag 31420: : , rating, attempt, optional pay). Submit it with POST /reviews (?dry_run=1 validates without storing) or the MCP tool submit_review, and read reviews with GET /reviews?domain= or get_reviews. Agents should use a fresh keypair dedicated to KlickyMart for pseudonymity and must never send a secret key. Reviews are signed claims; we verify signatures and format, not the opinion. Full guide: /llms.txt and /about#reviews. Spec: NIP-DRAFT.md → "422 as a merchant review".
Plain HTTP: GET /check?domain=example.com answers "can my agent shop here?" from the index only. It never starts a live crawl.
Crawler guardrails
- Honest user agent:
FloppyKitMapBot/0.2 (+https://klickymart.com/bot). No stealth plugins or fingerprint changes, no proxies or IP rotation. - robots.txt is read first; an error, 401, 403 or timeout means the whole site is treated as disallowed.
- If a site's terms ban automated access, the site is not crawled; only that finding is recorded, with a quote and URL.
- Public pages only: no login, no accounts, no orders, no payment details, no form submits beyond add-to-cart, no captcha solving. Blocks are recorded as findings.
- At least 2 s between requests, at most 10 pages per site.
- Customer reviews, Q&A and comments are never used as evidence. Every payment method in a record carries a quote and a URL; unknowns stay null.
Opt out: add User-agent: FloppyKitMapBot / Disallow: / to robots.txt. Details: https://klickymart.com/bot
Run the tests
cd crawler python3 -m venv .venv && . .venv/bin/activate pip install -r requirements.txt && python -m playwright install chromium python -m unittest discover -s tests -v
The tests use synthetic fixtures and make no network requests.
License
- Software (this repository, including the crawler and site Worker): MIT License.
- Crawl records (API JSON, NDJSON export, published directory data): CC0 1.0 Universal (public domain).
See also site/README.md.