Tcm Products MCP server

@pipeworx/tcm-products

Traditional Chinese Medicine (TCM) / Chinese Proprietary Medicine products licensed in Singapore — the Health Sciences Authority (HSA) "Listing of Licensed Chinese Proprietary Medicine Products" (~12,000 products), sourced from data.gov.sg, searchable by Chinese characters, pinyin (spaced or not), or English name.

Part of Pipeworx — an MCP gateway connecting AI agents to 1576+ live data sources.

An HSA licence is a safety and quality listing only — it is not evidence that a product treats any condition. The dataset carries no ingredients and no indications; it is a product inventory, and the tools describe it as one.

Tools

products matching a name in any script (板蓝根 / banlangen / Ban Lan Gen / SANJIN all work; the licensed name fields carry Chinese characters inline next to the romanised names, so matching needs no translation source). Filters are case-insensitive substring matches on their columns.

• tcmproductsearch(query?, company?, manufacturer?, country?, dosage_form?, limit?, offset?) —

aggregated with licensed-product counts and countries of manufacture; the supply-chain view of who makes licensed CPM products and where.

• tcmproductmanufacturers(country?, manufacturer?, limit?) — manufacturers

Why a separate pack, not tools on data-gov-sg

Fleet #1380. datagovsgquerydataset filters by exact column value, so it structurally cannot answer an alias query, and search_packs("traditional chinese medicine") found nothing TCM-shaped — the gap was discoverability. A pack whose name and descriptions say "Traditional Chinese Medicine" is what makes the catalog search find the data; two named tools buried in a general-purpose Singapore pack would not.

Auth

Keyless.

Data sources

the full export. One request returns all rows (limit=12000 → 11,977 rows, ~1.9 MB, verified 2026-09-08); the pack fetches it once per isolate and caches for 24h, because data.gov.sg rate-limits aggressively per caller (429s measured in fleet #1380) and the dataset updates roughly monthly.

• —

("q is invalid", verified 2026-09-08) — server-side text search does not exist on this surface, which is why matching is done in the Worker.

• CKAN's full-text q parameter is rejected by data.gov.sg

(managed by Health Sciences Authority, last updated 2026-08-07 at check time).

• Dataset metadata:

Licence

Singapore Open Data Licence v1.0 () — worldwide, perpetual, royalty-free, commercial use permitted with attribution. Verified 2026-09-08: the dataset page shows "Open Data Licence" and the licence text covers datasets on data.gov.sg. Attribution notice (included in every tool response): Contains information from "Listing of Licensed Chinese Proprietary Medicine Products" (Health Sciences Authority) from data.gov.sg, made available under the terms of the Singapore Open Data Licence version 1.0.

Quick Start

Add to your MCP client (Claude Desktop, Cursor, Windsurf, etc.):

{
  "mcpServers": {
    "tcm-products": {
      "url": "https://gateway.pipeworx.io/tcm-products/mcp"
    }
  }
}

What this endpoint actually serves

tools/list at https://gateway.pipeworx.io/tcm-products/mcp returns the tools in the table above plus the shared Pipeworx meta-tools — askpipeworx, discovertools, search_within, remember/recall and the rest of the gateway-wide set. So the tool count you see is larger than this table: a single-pack endpoint currently lists roughly 30 shared tools alongside the pack's own. The connection's initialize response states its exact scope, and is the authoritative answer for a given day.

Shortened. The full README is on GitHub.