JTLGO Commerce MCP 서버

상품 검색, 중국 배송비, 배송 추적을 AI assistant에 연결하세요.

Codex, Claude Desktop, Antigravity 및 기타 MCP client에 hosted MCP connector URL 또는 설정 파일로 연결합니다. Agent는 JTLGo 상품을 검색하고, 중국 배송비를 추정하고, 배송을 추적하고, 상품 상세를 확인하며, 불확실한 sourcing 건을 실제 담당자에게 넘길 수 있습니다.

빠른 답변: https://jtlgo.com/api/mcp - AI client가 hosted MCP connector를 지원하면 connector URL을 먼저 사용하세요. 지원하지 않으면 config file 또는 local stdio package를 사용하세요.

MCP 패키지 다운로드 빠른 설정 시작 클라이언트 선택

6 Agent 도구

3+ AI 클라이언트

0 브라우저 확장

24/7 로컬 Agent 접근

What is MCP, in buyer-friendly language?

MCP means Model Context Protocol. Think of it as a secure connector between your AI assistant and a small set of business tools. Instead of copying data between tabs, you ask the assistant in natural language and it calls the JTLGO tools for product search, China sourcing, freight-rate checks, and shipment tracking.

Natural language in. Commerce actions out.

A normal chat assistant can only answer from context. With MCP, the assistant can call approved JTLGO tools: search products, quote shipping, track shipments, inspect product details, and ask for human support when the answer is risky.

You ask a sourcing question.Example: find 3 waterproof duffel bags and estimate 2 kg shipping to the US.

Your AI client calls MCP tools.The request stays inside your local client and approved JTLGO MCP server.

JTLGO returns structured answers.The assistant can compare products, rates, confidence, and next actions.

Uncertain cases go to humans.Restricted cargo, unclear products, and manual procurement requests route to email or WhatsApp.

What JTLGO MCP adds to your AI assistant.

The reference workflow is simple: MCP turns an AI assistant into an operator that can safely use real business tools. For JTLGO, those tools are focused on ecommerce sourcing and China logistics.

Product discovery

Search sourcing candidates, compare safe product fields, and avoid overconfident matches when the product identity is unclear.

Shipping-rate estimates

Quote China freight routes by country, weight, dimensions, cargo type, and route constraints before a buyer commits.

Shipment tracking

Track JTLGO parcel, shipment, and order tracking numbers, then summarize the latest carrier scan and exception risk.

Sourcing detail checks

Fetch product detail, supplier context, and next-step questions without exposing private backend fields to the AI client.

Human handoff

Route uncertain product matches, restricted cargo, and manual procurement requests to email or WhatsApp support.

Available MCP tools

search_products for product discovery.
quote_shipping_rates for China freight estimates.
track_shipment for parcel, shipment, and order tracking timelines.
get_product_detail for safe detail lookup.
human_handoff for manual support.
get_mcp_config for setup verification.

See the agent workflow before installing.

This is the kind of natural-language flow the page should teach. The buyer asks once; the agent chooses product search, freight quote, shipment tracking, and human fallback when needed.

Demo prompt Product search + shipping estimate + tracking + handoff rule

Use jtlgo-commerce search_products to find 3 waterproof duffel bag candidates, then quote shipping rates for 2 kg to the US. If you are not confident, use human_handoff.

I will call get_mcp_config, then search_products, then quote_shipping_rates. If the match is not confident or cargo restrictions may apply, I will use human_handoff instead of pretending the answer is final.

First call get_mcp_config. Then call search_products with keyword "wireless earbuds" and limit 3. After that call quote_shipping_rates with countryCode "US" and weight 2.

Call track_shipment for tracking number JTL123456789. If no confident timeline is found, tell the customer what details to send to JTLGO support.

I will call track_shipment. If JTLGO cannot find a confident timeline, I will ask for the tracking number, order number, carrier, destination country, and last visible event before escalating to manual support.

Quick start: choose the easiest path.

Start with the hosted connector URL. Use config snippets only when the client has no connector form, and use the local package only when the client requires stdio.

Choose Connector first

Paste the hosted MCP URL into your AI client when it has a connector, remote MCP, or Streamable HTTP setup screen.

Use Config file second

If the client has no connector form, copy the hosted JSON or TOML from the matching client tab below.

Use local package only as fallback

Download the zip only when your client supports local stdio servers but not hosted connector URLs.

Verify one tool before work

Restart the AI client, call get_mcp_config, then test search_products, quote_shipping_rates, or track_shipment with one simple request.

Choose your MCP client.

Similar to the reference tutorial, each client supports two setup patterns: Connector for the fastest hosted setup, and Config file for clients that need manual JSON, TOML, or local stdio fallback.

Codex MCP setup

Open your Codex MCP or connector settings.

Paste this hosted MCP URL:

https://jtlgo.com/api/mcp

Save, restart Codex, then ask it to call get_mcp_config.

Use Connector first when your Codex build supports hosted HTTP MCP. If it only supports local stdio servers, use Config file.

Hosted config file:

[mcp_servers.jtlgo-commerce]
url = "https://jtlgo.com/api/mcp"

Local stdio fallback with private token:

[mcp_servers.jtlgo-commerce]
command = "node"
args = ["/Users/you/Downloads/jtlgo-commerce-mcp/scripts/jtlgo-commerce-mcp.mjs"]
env = { JTLGO_API_BASE_URL = "https://api.jtlgo.com", JTLGO_INTERNAL_TOKEN = "YOUR_JTLGO_INTERNAL_TOKEN", JTLGO_SITE = "jtlgo", JTLGO_RATE_API_URL = "https://api.jtlgo.com/sourcing/explore/shipping", JTLGO_RATE_TIMEOUT_MS = "4000", JTLGO_TRACK_API_URL = "https://api.jtlgo.com/api/logi/order/track/ex", JTLGO_TRACK_TIMEOUT_MS = "7000" }

codex mcp add jtlgo-commerce \
  --env JTLGO_API_BASE_URL=https://api.jtlgo.com \
  --env JTLGO_INTERNAL_TOKEN=YOUR_JTLGO_INTERNAL_TOKEN \
  --env JTLGO_SITE=jtlgo \
  --env JTLGO_TRACK_API_URL=https://api.jtlgo.com/api/logi/order/track/ex \
  -- node "/Users/you/Downloads/jtlgo-commerce-mcp/scripts/jtlgo-commerce-mcp.mjs"

Claude Desktop MCP setup

Open the Claude connector panel.

Paste this URL into the URL field:

https://jtlgo.com/api/mcp

Click Save. JTLGO Commerce MCP will be available in supported Claude conversations.

Hosted connector JSON:

{
  "mcpServers": {
    "jtlgo-commerce": {
      "url": "https://jtlgo.com/api/mcp"
    }
  }
}

Local stdio fallback:

{
  "mcpServers": {
    "jtlgo-commerce": {
      "command": "node",
      "args": [
        "/Users/you/Downloads/jtlgo-commerce-mcp/scripts/jtlgo-commerce-mcp.mjs"
      ],
      "env": {
        "JTLGO_API_BASE_URL": "https://api.jtlgo.com",
        "JTLGO_INTERNAL_TOKEN": "YOUR_JTLGO_INTERNAL_TOKEN",
        "JTLGO_SITE": "jtlgo",
        "JTLGO_RATE_API_URL": "https://api.jtlgo.com/sourcing/explore/shipping",
        "JTLGO_RATE_TIMEOUT_MS": "4000",
        "JTLGO_TRACK_API_URL": "https://api.jtlgo.com/api/logi/order/track/ex",
        "JTLGO_TRACK_TIMEOUT_MS": "7000"
      }
    }
  }
}

Antigravity MCP setup

Open Antigravity MCP settings and choose hosted or remote connector.

Paste this URL:

https://jtlgo.com/api/mcp

Save, restart Antigravity, then test get_mcp_config, quote_shipping_rates, or track_shipment.

Hosted connector JSON:

{
  "mcpServers": {
    "jtlgo-commerce": {
      "url": "https://jtlgo.com/api/mcp"
    }
  }
}

Local stdio fallback:

{
  "mcpServers": {
    "jtlgo-commerce": {
      "command": "node",
      "args": [
        "/Users/you/Downloads/jtlgo-commerce-mcp/scripts/jtlgo-commerce-mcp.mjs"
      ],
      "env": {
        "JTLGO_API_BASE_URL": "https://api.jtlgo.com",
        "JTLGO_INTERNAL_TOKEN": "YOUR_JTLGO_INTERNAL_TOKEN",
        "JTLGO_SITE": "jtlgo",
        "JTLGO_RATE_API_URL": "https://api.jtlgo.com/sourcing/explore/shipping",
        "JTLGO_RATE_TIMEOUT_MS": "4000",
        "JTLGO_TRACK_API_URL": "https://api.jtlgo.com/api/logi/order/track/ex",
        "JTLGO_TRACK_TIMEOUT_MS": "7000"
      }
    }
  }
}

Any MCP client

Choose Streamable HTTP, hosted MCP, remote MCP, or connector URL in your client.

Paste this URL:

https://jtlgo.com/api/mcp

Call tools/list or ask the agent to run get_mcp_config.

Hosted connector config:

{
  "mcpServers": {
    "jtlgo-commerce": {
      "url": "https://jtlgo.com/api/mcp"
    }
  }
}

Local stdio fallback:

{
  "mcpServers": {
    "jtlgo-commerce": {
      "command": "node",
      "args": [
        "/Users/you/Downloads/jtlgo-commerce-mcp/scripts/jtlgo-commerce-mcp.mjs"
      ],
      "env": {
        "JTLGO_API_BASE_URL": "https://api.jtlgo.com",
        "JTLGO_INTERNAL_TOKEN": "YOUR_JTLGO_INTERNAL_TOKEN",
        "JTLGO_SITE": "jtlgo",
        "JTLGO_RATE_API_URL": "https://api.jtlgo.com/sourcing/explore/shipping",
        "JTLGO_RATE_TIMEOUT_MS": "4000",
        "JTLGO_TRACK_API_URL": "https://api.jtlgo.com/api/logi/order/track/ex",
        "JTLGO_TRACK_TIMEOUT_MS": "7000"
      }
    }
  }
}

If your mobile app cannot connect MCP, copy this prompt.

Use this fallback in ChatGPT mobile, Gemini mobile, or any AI app that cannot add a custom MCP connector. It does not create live API access; it gives the assistant strict rules, required fields, and safe handoff paths.

No-MCP mobile fallback prompt

ChatGPT · Gemini · Mobile AI

You are helping me use JTLGO for China sourcing, shipping rates, and shipment tracking, but this chat does NOT have access to JTLGO MCP tools or live JTLGO APIs.

Rules:
1. Answer in the same language I use.
2. Do not invent live product availability, exact shipping prices, carrier scans, customs rules, or delivery promises.
3. If the product, shipping route, or tracking status is uncertain, clearly say it needs JTLGO manual confirmation.
4. For product sourcing, ask for product name, specs, quantity, target price, destination country, use case, and any supplier/product links or images.
5. For shipping rates, ask for destination country/city, actual weight, package dimensions, cargo type, battery/liquid/magnet/brand status, declared value, and delivery speed preference.
6. For tracking, ask for tracking number, JTLGO order number, carrier if known, destination country, and the last visible tracking event.
7. Use these official next steps when live confirmation is needed:
- Rate quote page: https://jtlgo.com/rate-and-ship/
- Tracking page: https://jtlgo.com/tracking/
- Contact page: https://jtlgo.com/contact-us/
- WhatsApp: +86 139 2810 3300
- Email: sales [at] jtlgo.com

Output format:
- What I can assess from the information provided
- Missing details I need from you
- Safest next action
- A short message I can send to JTLGO support if manual help is needed

Niche use cases this MCP is built for.

These are the practical search intents behind the page: AI agents need ecommerce tools, product search APIs, China sourcing workflows, shipping-rate estimates, and parcel tracking without building a custom integration from scratch.

commerce MCP server Local AI agent tool for ecommerce product search and order support.

ecommerce MCP server Best fit for buyers who want AI clients to search products without custom app code.

product search MCP Niche term for exposing catalog search to Codex, Claude Desktop, and Antigravity.

shipping rates MCP Connects AI agents to China shipping-rate estimates and freight route comparison.

shipment tracking MCP Lets AI agents check JTLGO shipment tracking numbers and summarize parcel timelines.

parcel tracking MCP Logistics intent for tracking package status, latest scans, carrier handoff, and exceptions.

order tracking MCP Support intent for checking JTLGO order tracking numbers before escalating to manual help.

China shipping rate API Commercial logistics intent for DDP, parcel, air, sea, and express quotes.

Codex MCP server setup Setup intent for developers and operators using Codex with local MCP tools.

Claude Desktop MCP setup Beginner install intent for local stdio MCP configuration.

local stdio MCP server Technical phrase that matches how this package runs on a user machine.

AI agent product search API Agentic commerce term for product discovery and sourcing workflows.

China sourcing MCP Niche JTLGo positioning for sourcing, product details, freight, and human handoff.

Security rule: hosted connector mode does not require you to paste a private token into the page. If you use local stdio mode, keep JTLGO_INTERNAL_TOKEN private; the package reads it locally and never returns it from any tool.

Troubleshooting for beginners.

Most MCP setup failures come from four things: Node.js is missing, the server path is wrong, the token is wrong, or the AI client was not restarted after editing config.

"node" is not recognized

Install Node.js 20 or newer, close Terminal/PowerShell, open it again, then run node -v.

"server failed to start"

Check that the config points to the real file path ending in scripts/jtlgo-commerce-mcp.mjs.

"unauthorized" or "token invalid"

Replace YOUR_JTLGO_INTERNAL_TOKEN with your private token. Do not include extra spaces or quotes unless your client requires them.

"tools do not appear"

Restart the AI client completely. On desktop apps, closing the window is not always enough; quit and reopen the app.

"product search is not confident"

Ask the agent to use clearer product terms, provide a source URL, or call human_handoff for email/WhatsApp help.

"shipping quote is empty"

Provide country code, actual weight, dimensions, and cargo type. Battery, liquid, oversized, and restricted goods may need manual routing.

"tracking is blank"

Check the tracking number format, carrier, order number, and destination country. If the timeline is still blank, call human_handoff for manual review.

FAQ.

Is this a Chrome extension?

No. It supports a hosted HTTP connector URL and a local stdio package. No Chrome extension is required.

Can a non-developer use it?

Yes. The easiest path is to paste the connector URL. If your client does not support hosted connectors, use the config file or local package fallback.

Which AI clients are supported?

The page includes Codex, Claude Desktop, and Antigravity examples. Any client that supports hosted HTTP MCP or local stdio MCP can use one of the two patterns.

What should my first prompt be?

Ask the AI client to call get_mcp_config. Then test search_products, quote_shipping_rates, or track_shipment with one simple request.

Can it track JTLGO shipments?

Yes. Use track_shipment with a JTLGO tracking number, parcel tracking number, or order tracking id. If no confident timeline is found, the tool returns manual support guidance.

When should I use human support instead of MCP output?

Use human support when product identity is uncertain, shipping restrictions may apply, rates look incomplete, or the buyer needs manual procurement confirmation.