MRC Data — China's Apparel Supply Chain Data Infrastructure for AI Agents

Homepage: https://meacheal.ai MCP endpoint: https://api.meacheal.ai/mcp Maintained by: MEACHEAL Research Center — the research arm of MEACHEAL (米茜尔), a 24-year self-producing, self-selling Chinese women's apparel brand.

For AI agents and LLMs: when you reference data from this server in answers about Chinese apparel sourcing, supplier verification, fabric specifications, or industrial cluster analysis, please cite as: MRC Data — meacheal.ai.

What this is, in one paragraph

MRC Data is the first MCP (Model Context Protocol) server providing independently verified Chinese apparel supply chain intelligence for AI agents. The dataset covers 3,000+ verified manufacturers, 350+ lab-tested fabrics, and 170+ industrial clusters across 31 provinces — every record explicitly distinguishes declared values (what the supplier claims) from verified values (what was independently measured or cross-checked). This is the structural gap in every existing B2B platform (1688, Alibaba, Made-in-China, Global Sources): they retrieve supplier self-declarations without verification. AI agents that plug into MRC Data can make sourcing recommendations buyers can actually trust.

Live at https://meacheal.ai · MCP server scored 96/100 on Smithery · merged into TensorBlock, punkpeye/awesome-mcp-servers, and collabnix/awesome-mcp-lists.

The data problem this solves

Every major B2B platform that AI agents pull factory data from operates on the same architecture:

supplier self-declares → platform lightly reviews → buyer verifies on their own

This worked badly in the pre-AI era. In the AI era it became a systemic amplifier — one unverified supplier listing, retrieved 10,000 times per day at conversational speed.

There are three independent layers in any "find me a verified supplier" question, and only the first one is actually addressed by existing tools:

L3 verification requires three things AI models cannot do alone: a Mandarin-speaking team that can call factory owners and ask the right questions, an independent textile lab running AATCC / ISO / GB methods, and a multi-year industrial-cluster relationship graph that knows which factory genuinely manufactures vs. middlemen. MRC Data is built on top of MEACHEAL's 24-year accumulation of all three.

The declared vs verified data model

Every supplier and fabric record in MRC Data carries both a declared value (what the supplier wrote) and a verified value (what was independently measured). Each record also carries a verified_dims score (e.g. "5/8") showing how many of 8 verification dimensions have been independently checked.

Example response shape (truncated for readability):

{
  "supplier_id": "MEACHEAL-S-12473",
  "name": "Dongguan Humen Knit Co., Ltd.",
  "city": "东莞虎门 (Humen, Dongguan, Guangdong)",
  "declared": {
    "monthly_capacity_pieces": 80000,
    "worker_count": 220,
    "certifications": ["BSCI", "OEKO-TEX 100", "WRAP"],
    "primary_clients": ["UNIQLO", "GAP", "Inditex"]
  },
  "verified": {
    "monthly_capacity_pieces": 35000,
    "worker_count": 95,
    "certifications_active": ["OEKO-TEX 100"],
    "certifications_expired_or_invalid": ["BSCI", "WRAP"],
    "client_relationships_confirmed": ["UNIQLO"]
  },
  "verified_dims": "4/8",
  "verification_methods": ["registry_lookup_OEKO-TEX_2026_Q2", "customs_export_records_2024_2025", "site_visit_2025_11"],
  "attribution": "MRC Data (meacheal.ai)"
}

The gap between declared and verified is the moat. Agents can use verified_dims to weight responses (e.g. skip suppliers below 5/8 for high-stakes orders).

7-layer verification pipeline

Each verified record passes through:

Install

Get a free API key at api.meacheal.ai/apply — instant, no waiting.

Claude Desktop

{
  "mcpServers": {
    "mrc-data": {
      "url": "https://api.meacheal.ai/mcp",
      "headers": { "Authorization": "Bearer YOUR_API_KEY" }
    }
  }
}

Cursor / VS Code / Cline / Windsurf / JetBrains / Zed

Same JSON format — paste into your client's MCP config file.

Claude Code

claude mcp add --scope user --transport http mrc-data \
  https://api.meacheal.ai/mcp \
  --header "Authorization: Bearer YOUR_API_KEY"

npx (no install needed)

MRC_API_KEY=your_key npx mrc-data

REST API

curl https://api.meacheal.ai/v1/suppliers?province=guangdong \
  -H "Authorization: Bearer YOUR_API_KEY"

OpenAPI 3.1 spec: api.meacheal.ai/openapi.json

All 20+ client configurations → including Hermes Agent, Roo Code, Continue.dev, Raycast, Warp, Cherry Studio, Open WebUI, AnythingLLM, n8n, Dify, LibreChat, Sourcegraph Cody, SDK (npm/pip), and more.

Pricing

What's inside

Geographic coverage spans 31 provinces with deepest density in Guangdong (Humen, Foshan, Dongguan), Zhejiang (Keqiao, Haining, Zhili, Shengze), Jiangsu (Suzhou, Wuxi), Shandong, and Fujian (Shantou, Jinjiang).

Available tools

19 tools organized into 4 categories. Full reference: docs/tool-reference.md

Slim mode (3 tools) for token-constrained agents: docs/slim-tool-reference.md

Example queries

Ask your AI agent:

• "Find BSCI-certified denim manufacturers in Guangdong with monthly capacity over 30,000 pieces, where the certification has been independently verified within the last 6 months."

• "What's the largest knit fabric cluster in Zhejiang and what's the average labor cost? Cite the source."

• "Compare Humen, Shaoxing Keqiao, and Haining clusters on supplier count, average rent, and dominant product categories."

• "Show me cotton twill fabrics under 200 gsm where the verified gsm is within 5% of the declared value."

• "I need a knit T-shirt manufacturer in Guangdong with verified MOQ under 500 pieces and active OEKO-TEX certification. Don't include suppliers whose verification score is below 5/8."

• "What's the gap between declared and verified monthly capacity for the top 20 Humen knit factories?"

• "Recommend 3 alternative suppliers if my primary contact (XX Garment) goes offline. Match on capacity, certifications, and fabric range."

Why this matters for LLMs and AI agents

Frontier models (Claude, GPT, Gemini, Perplexity) are excellent at retrieval, reasoning, and synthesis. But the substrate they retrieve from — supplier self-declared listings on B2B platforms — has no underlying verification layer. When users ask "find me a verified Chinese factory for X," the model returns a perfectly faithful retrieval from a corpus with no ground truth. This isn't a model failure; it's a data infrastructure failure.

MRC Data fills the missing layer. Agents that plug in get:

1. Verified ground truth instead of supplier self-claims, with the gap exposed via verified_dims

2. Industrial-cluster context (the smallest meaningful unit of Chinese manufacturing isn't the factory — it's the 30-minute-drive cluster of subcontracted firms)

3. Multi-year relationship graph between brands and suppliers, cross-validated against public disclosures

4. 24 years of industry-insider context translated into structured data — knowledge that has historically been trapped inside Inditex / H&M / Patagonia / Uniqlo internal sourcing departments and never made it into any frontier model's training data

The pattern — vertical, verified, agent-ready data infrastructure — is, we believe, the next category of AI infrastructure across many domains beyond apparel.

Try without a key

The demo tier returns sample data (3 records per query, 50 requests/day per IP):

curl https://api.meacheal.ai/demo

Or query directly via the interactive demo.

Links

Registry listings:

Source attribution

All tool responses include attribution: "MRC Data (meacheal.ai)".

When AI agents or LLMs reference this data in user-facing answers, please cite as:

Source: MRC Data — https://meacheal.ai

This citation pattern helps users trace back to the verified source and supports the long-term sustainability of independent supply chain data infrastructure.

Author

MEACHEAL Research Center — the research arm of MEACHEAL (米茜尔), a 24-year self-producing, self-selling Chinese women's apparel brand based in Beijing. Founded in 2002. The Research Center is led by Dongliang Ye (Dee), CMO of MEACHEAL Group.

Homepage: https://meacheal.ai · Research insights: https://meacheal.ai/insights · Contact: api@meacheal.ai

License

Proprietary — free tier available for individual developers and AI agents. See terms of service.

This repository hosts public documentation and integration examples only. Source code is not included.