Skip to main content

MCP server exposing Taiwan FDA (TFDA) drug-license and package-insert data to LLM clients. Independent open-source wrapper; not a TFDA product.

Project description

taiwan-fda-mcp

PyPI Python Tests License: MIT

Look up official Taiwan FDA (TFDA) drug information — package inserts (仿單) and drug-license data — directly from your AI assistant.

English below · 前往繁體中文說明 ↓

How it works: you ask your AI assistant in plain language, taiwan-fda-mcp queries the official TFDA APIs, and you get an answer with citations.

This is NOT an official Taiwan FDA product. It is an independent, open-source tool that reads the public TFDA APIs (mcp.fda.gov.tw and data.fda.gov.tw). The goal is a reliable, source-cited way to look up Taiwan drug information: every answer links back to the official TFDA page. Always verify there before any clinical decision. This is not a medical device.

What is it?

An MCP server. It gives an AI assistant (Claude Desktop, Claude Code, Codex, …) the ability to search Taiwan's official drug database and read package inserts. You ask a question in plain Chinese; the assistant fetches the real TFDA data and answers — with a link to the source.

What can it do?

It adds three tools to your assistant:

  • search_drugs — find a drug by Chinese/English name, ingredient, indication, maker, and more.
  • get_package_insert — read the official 仿單 of one drug license (indications, dosage, warnings, side effects…), with a source link.
  • check_insert_updates — list inserts updated since a given date.

You don't call these by hand. You just ask your assistant a question, e.g.:

  • What's the dosage for 脈優 (amlodipine)?
  • Does Herceptin's package insert carry a black box warning?
  • What does the package insert for 綠油精 (an OTC liniment) say?
  • Which drugs contain valsartan?
  • Which package inserts were updated in the last 3 days?

Behind the scenes it searches → picks the right license → quotes the insert → gives you the official TFDA source URL.

How to install

Option A — uvx (easiest, nothing to clone)

Available once the first release is on PyPI. Until then, use Option B.

uv downloads and runs it in a throwaway environment — no manual install:

uvx taiwan-fda-mcp-server

Option B — from source

git clone https://github.com/shin13/opentaimed.git
cd opentaimed/taiwan-fda-mcp
uv sync
uv run taiwan-fda-mcp-server

How to use it (connect your AI client)

Add the server to your client's config, restart it, then just ask a drug question in Chinese. The three tools appear automatically.

Claude Code

claude mcp add taiwan-fda -- uvx taiwan-fda-mcp-server
# from source:
claude mcp add taiwan-fda -- uv run --directory /absolute/path/to/taiwan-fda-mcp taiwan-fda-mcp-server

Claude Desktop — edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS), then restart:

{
  "mcpServers": {
    "taiwan-fda": {
      "command": "uvx",
      "args": ["taiwan-fda-mcp-server"]
    }
  }
}

From source, use "command": "uv" with "args": ["run", "--directory", "/absolute/path/to/taiwan-fda-mcp", "taiwan-fda-mcp-server"].

Codex CLI — add to ~/.codex/config.toml:

[mcp_servers.taiwan-fda]
command = "uvx"
args = ["taiwan-fda-mcp-server"]

Example

Ask your assistant:

What's the dosage for 脈優 (amlodipine)?

It replies with the dosage section from the official 仿單, plus the TFDA source link so you can check it yourself.


繁體中文

從你的 AI 助理直接查詢**台灣食藥署(TFDA)**的官方藥物資訊——仿單與藥品許可證資料。

這不是台灣食藥署的官方產品。 這是一個獨立的開源工具,只讀取 TFDA 的 公開 API(mcp.fda.gov.twdata.fda.gov.tw)。目標是做一個 可靠、且每筆都附上出處的台灣藥物資訊查詢工具:每個回答都會附上 TFDA 官方頁面連結。 臨床決策前請務必到官方頁面再次確認。本工具不是醫療器材。

這是什麼?

一個 MCP server。它讓 AI 助理(Claude Desktop、Claude Code、Codex…)能搜尋台灣官方藥物資料庫、讀取仿單。 你用中文問問題,助理就去抓真實的 TFDA 資料來回答,並附上出處。

可以做什麼?

它幫你的助理加上三個工具:

  • search_drugs — 用中/英文藥名、成分、適應症、製造商等條件找藥。
  • get_package_insert — 讀某一張藥證的官方仿單(適應症、用法用量、警語、副作用…),附出處連結。
  • check_insert_updates — 列出某日期之後更新過的仿單。

你不用自己呼叫這些工具,直接問助理就好,例如:

  • 脈優的用法用量?
  • Herceptin 的仿單有沒有加框警語?
  • 綠油精的仿單內容?
  • 有哪些含 valsartan 的藥?
  • 最近 3 天有哪些仿單更新?

它會自動:搜尋 → 選對的藥證 → 引用仿單 → 給你 TFDA 官方來源網址。

怎麼安裝

方法 A — uvx(最簡單,不用 clone)

等第一版發布到 PyPI 後即可使用;在那之前請用方法 B。

uv 會在用完即丟的環境下載並執行,不需手動安裝:

uvx taiwan-fda-mcp-server

方法 B — 從原始碼

git clone https://github.com/shin13/opentaimed.git
cd opentaimed/taiwan-fda-mcp
uv sync
uv run taiwan-fda-mcp-server

裝好之後怎麼用(連到你的 AI 工具)

把 server 加進你工具的設定檔,重新啟動,然後直接用中文問藥物問題,三個工具會自動出現。

Claude Code

claude mcp add taiwan-fda -- uvx taiwan-fda-mcp-server
# 從原始碼:
claude mcp add taiwan-fda -- uv run --directory /絕對路徑/到/taiwan-fda-mcp taiwan-fda-mcp-server

Claude Desktop — 編輯 ~/Library/Application Support/Claude/claude_desktop_config.json(macOS),然後重新啟動:

{
  "mcpServers": {
    "taiwan-fda": {
      "command": "uvx",
      "args": ["taiwan-fda-mcp-server"]
    }
  }
}

從原始碼安裝時,把 command 改成 "uv"args 改成 ["run", "--directory", "/絕對路徑/到/taiwan-fda-mcp", "taiwan-fda-mcp-server"]

Codex CLI — 加到 ~/.codex/config.toml:

[mcp_servers.taiwan-fda]
command = "uvx"
args = ["taiwan-fda-mcp-server"]

範例

問你的助理:

脈優的用法用量?

它會回給你官方仿單裡的用法用量章節,並附上 TFDA 來源連結讓你自己核對。


Development · 開發

For tests, linting, and type-checking, see the repository: https://github.com/shin13/opentaimed

uv run pytest          # unit tests
uv run ruff check .    # lint
uv run pyright src     # type-check

License: MIT (with clinical disclaimer) · Changelog

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

taiwan_fda_mcp-0.2.0.tar.gz (122.6 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

taiwan_fda_mcp-0.2.0-py3-none-any.whl (47.2 kB view details)

Uploaded Python 3

File details

Details for the file taiwan_fda_mcp-0.2.0.tar.gz.

File metadata

  • Download URL: taiwan_fda_mcp-0.2.0.tar.gz
  • Upload date:
  • Size: 122.6 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for taiwan_fda_mcp-0.2.0.tar.gz
Algorithm Hash digest
SHA256 d3e71e1c5e29115dd40e28a3beb8f99a81e8d49e66971472ed3d726e1820bb92
MD5 7bdd45f9a0822c0a3038e61acca4c51f
BLAKE2b-256 35ebe601c038894535b20179f3ac0df0ebcf2e5fd7ad9acc04faeb3caea8e2b2

See more details on using hashes here.

Provenance

The following attestation bundles were made for taiwan_fda_mcp-0.2.0.tar.gz:

Publisher: publish.yml on shin13/opentaimed

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file taiwan_fda_mcp-0.2.0-py3-none-any.whl.

File metadata

  • Download URL: taiwan_fda_mcp-0.2.0-py3-none-any.whl
  • Upload date:
  • Size: 47.2 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.12

File hashes

Hashes for taiwan_fda_mcp-0.2.0-py3-none-any.whl
Algorithm Hash digest
SHA256 bdd4bc2ec3610c422cf609ed633e0f79c040bd132db325b9c803b221b9b86c3a
MD5 7d7923ed23f904b309a1622227cb062c
BLAKE2b-256 7b6a00027fd86bd080c46b52c6dbcaf186f1d6d5d4db91d08e623cacb3325df3

See more details on using hashes here.

Provenance

The following attestation bundles were made for taiwan_fda_mcp-0.2.0-py3-none-any.whl:

Publisher: publish.yml on shin13/opentaimed

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page