The full upstream README, mirrored here for reference. Install config, tool schemas, adoption signals, and an original overview live on the China Pharma listing page.
Biotech and pharmaceutical licensing deals and drug approvals involving China, read out of SEC 8-K and HKEX disclosure announcements. Keyless.
Part of Pipeworx — an MCP gateway connecting AI agents to 1683+ live data sources.
china_licensing_deals(query?, since?, until?, limit?) — deals where a Chinese company
out-licensed an asset to a Western partner, or a Western company licensed rights into Greater
China. Returns both companies, the territory clause, upfront and milestone amounts, royalty
language, and the filing it was announced in.nmpa_updates(section?, limit?) — official NMPA announcements in English: policy, guidance,
import rules, clinical-trial guideline updates. Sections: news, drugs, devices, cosmetics.
NOT the approvals register — see below.nmpa_drug_approvals(month? | since?/until?, query?, limit?) — specific drugs approved,
conditionally approved, or given an expanded indication by NMPA, read out of HKEX (Hong Kong
Exchange) disclosure filings. Returns the filing headline, company, approval date, a best-effort
category, and the filing PDF.nmpa_drug_approval(id? | name?) — one approval record by the id from
nmpa_drug_approvals, or by drug/company name search.csrc_penalties(since?, page?, limit?) — administrative penalty decisions (行政处罚决定书)
from the CSRC, China's securities regulator, newest first from csrc.gov.cn's own JSON feed.
Each row carries the decision number, publication date, the first named party, a memo excerpt
with the remaining parties, and the link to the full decision. Securities enforcement —
disclosure fraud, insider trading, audit failures — not drug regulation.nmpa_updates is not, and where the approvals actually come fromnmpa_updates is not the drug-approval register, and neither — contrary to what it looks
like on a first read — is english.nmpa.gov.cn/drugs.html. Probing it live (2026-09-07) rather
than trusting the page name: every section (drugs, news, policies,
regulatoryinformation) is a paginated feed of administrative-policy notices ("Announcement on
Strengthening Supervision of Contract Manufacturing", "Policy Interpretation of..."), the newest
item under drugs is dated 2026-04-14 — five months stale against a September run — and not one
entry across any section names a specific drug as approved. The Chinese-language NMPA site (all
paths, not just /datasearch/, confirmed domain-wide this run) and CDE both still answer HTTP
412/202 with a 瑞数/Riverdata JS-obfuscation challenge to a plain non-Cloudflare client — the
2026-08-11 finding holds. Reading either needs a headless browser, which a stateless pack cannot
be.
What actually carries approval events, keyless and current: any HKEX-listed Chinese biotech is
required to disclose an NMPA approval as a same-day regulatory filing, and HKEXnews' public
title-search JSON endpoint (titleSearchServlet.do) serves those without a key — the same trick
china_licensing_deals already uses one exchange over (SEC 8-Ks instead of HKEX filings): the
curated register isn't public, but the disclosure layer is. nmpa_drug_approvals and
nmpa_drug_approval are built on that.
titleSearchServlet.do ignores from/to when searchType=1 and title is set. Verified
by requesting three different multi-month windows (Jan–Sep 2026, July 2026 only, a single day)
and getting back the identical rolling set of the ~17 most-recent NMPA-titled filings every
time. This pack fetches that one unbounded window and does its own date filtering
client-side — since/until/month outside the reachable window return found: false with a
hint explaining why, not a silently-empty "no approvals that month".classifyApproval()
excludes them by keyword (IND, clinical trial, accept(ed|ance), designation,
clearance, submitted) before accepting a title as a marketing approval.name_cn is deliberately null. HKEXnews' English title search returns English-language
headlines only; the paired Chinese-language disclosure is a different NEWS_ID and isn't
reliably matchable without another lookup, and the PDFs aren't parsed. Left explicit rather than
guessed.scope_note.None. SEC asks for a contactable User-Agent and 403s a generic one; the pack sends one.
Deal databases (Citeline, Evaluate, Cortellis, DealForma) are all subscription products. But any US-listed party to a licensing deal announces it in an 8-K, and the press-release exhibit contains the counterparty, the territory and the money — so the announcement layer is public even though the curated database is not.
The scoping is the whole trick. "license agreement" AND "China" returns Yum China and YUM
Brands ahead of any biotech, because fast-food franchising uses the same words. Requiring deal
vocabulary instead — "upfront payment" AND "milestone" AND "China" — returned 24 filings over
three months, every one of them a pharma company. The domain filter is the DEAL language, not the
word "pharma".
coverage_note.$1) is deliberately rejected —
no licensing deal is denominated in single dollars, and reporting one would be a wrong number.query to bring it into range.csrc_penalties reads the searchList JSON endpoint behind
csrc.gov.cn/csrc/c101928/common_list.shtml (channel 28de6b87eda140cb93de4dd10d11867d,
行政处罚). Unlike NMPA's Chinese site, this endpoint carries no JS-challenge WAF — a plain
request with a Referer header answers 200 (verified 2026-09-07). Every decision is titled with
the regulator's fixed boilerplate, so the useful fields are parsed from memo: the decision
number (〔2026〕42号) and the 当事人 (named parties) block. The linked content.shtml pages are
HTML prose and are deliberately not fetched — the tool is a dated feed with links, not a
parsed penalty database. _rangeTimeGte filters server-side but is loose at the day boundary
(a since=2026-08-01 request came back with a 2026-07-31 row), so since is re-applied
client-side to the returned rows.
china_licensing_deals: https://efts.sec.gov/LATEST/search-index (EDGAR full-text search,
8-K forms) and the filing bodies under https://www.sec.gov/Archives/edgar/data/.nmpa_updates: https://english.nmpa.gov.cn/ (news, drugs, medicaldevices, cosmetics
sections).nmpa_drug_approvals / nmpa_drug_approval: https://www1.hkexnews.hk/search/titleSearchServlet.do
(HKEXnews public title search) and the linked filing PDFs under https://www1.hkexnews.hk/.csrc_penalties: https://www.csrc.gov.cn/searchList/28de6b87eda140cb93de4dd10d11867d
(the 行政处罚 channel's JSON list endpoint).Add to your MCP client (Claude Desktop, Cursor, Windsurf, etc.):
tools/list at https://gateway.pipeworx.io/china-pharma/mcp returns the tools in the table
above plus the shared Pipeworx meta-tools — ask_pipeworx,
discover_tools, 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.
This is deliberate, not multiplexing by accident. The meta-tools are what let a
scoped connection answer a question this pack does not cover — via
ask_pipeworx, which routes across the whole catalog — without you adding a
second MCP server. There is currently no way to mount a pack endpoint without
them; if the extra schemas cost you more context than the routing is worth,
connect to the full gateway once rather than to several pack endpoints.
Or connect to the full Pipeworx gateway to get every pack's tools listed directly, instead of just this one's:
Both URLs reach the same gateway and the same 1683+ data sources. The
only difference is which pack's tools are listed directly; ask_pipeworx
reaches all of them from either one.
No account needed for the first calls. Inspect any tool: GET https://gateway.pipeworx.io/v1/tools/china_licensing_deals. Find one: POST https://gateway.pipeworx.io/v1/tools/search_packs with {"query":"..."}.
This package also runs as a local stdio MCP server — no Pipeworx account, no gateway round-trip:
Or run it directly to confirm it starts:
It speaks MCP over stdin/stdout and answers initialize/tools/list/tools/call
for only this pack's tools — none of the shared meta-tools the gateway
connection above adds. Same source, same tools, no ask_pipeworx routing.
Instead of calling tools directly, you can ask questions in plain English — this works on the pack endpoint above as well as on the full gateway:
The gateway picks the right tool and fills the arguments automatically.
MIT