Twinkle HubTwinkle Hub
Sign in

📌 2026-05-13 added: Nationwide real-estate transaction registry consolidated (3 datasets · 6,030,000+ rows)

View full changelog →

🎉 Free during alpha · sign up now

Taiwan data MCP Hub

MCP is the open protocol for AI agents — Claude, Cursor, OpenClaw, Hermes, your own agent loop. Twinkle Hub bundles Taiwan data sources behind a single MCP endpoint — set your client up once, AI can query datasets, run SQL, and call tools.

Sign up freeSee pricing

✨ One-paste install

Paste this to Claude — it'll set up the MCP for you

Add Twinkle Hub to my Claude Desktop. Edit
~/Library/Application Support/Claude/claude_desktop_config.json
(Windows: %APPDATA%\Claude\claude_desktop_config.json).

Merge in this block:
{
  "mcpServers": {
    "twinkle-hub": {
      "type": "http",
      "url": "https://api.twinkleai.tw/mcp/",
      "headers": {
        "Authorization": "Bearer <I'll paste from https://hub.twinkleai.tw/dashboard>"
      }
    }
  }
}

If an mcpServers block already exists, merge — don't overwrite.
Then ask me to restart Claude and list the 5 new opendata-* tools.

Also works in Cursor / Claude Code / any MCP client. Full install paths in /docs.

🎁 Free during alpha: all tools, no monthly cap, no rate limits (within fair-use).

52,960

Datasets

Layer 1: OpenData

20

Domains

auto-classified

42

TW utility tools

Layer 2: local-first

8

BYO connectors

Layer 3: client-side

Compatible with

Twinkle Hub speaks MCP. So does every client below.

  • Claude
  • Cursor
  • Gemini
  • OpenAI
  • Ollama
  • Cline

Plus your own — MCP is an open protocol.

Recent updates

The data keeps growing

We sync from ministries and licensed third-party sources every day. The three most recent additions are below.

View full changelog
2026-05-13v1.13.0

Nationwide real-estate transaction registry consolidated

rows added

+6,030,000

datasets

3

  • lvr-trades 4.75M rows of sales / lvr-presale 554k rows of presale incl. rescissions / lvr-rentals 726k rows of rentals
  • Coverage: 22 counties × 55 quarters (2012-Q3 → 2026-Q1)
  • Preserves all 33-35 original Traditional-Chinese columns + 3 enrichment fields (city / season / iso_trade_date)

Try asking:

  • Top 5 highest-priced property sales in Xinyi district 2024
  • Property sales count in Zhongli district, Taoyuan, last year
2026-05-13v1.12.0

Government procurement greatly expanded

rows added

+27,189

datasets

1

  • pcc-tender grew 135k → 162,189 rows (11 years of complete history from 2015-04)
  • Schema expanded 9 → 19 columns: procurement type, tender method, award price, agency phone/address, vendor address, losing bidders, more
  • Sourced from web.pcc.gov.tw (Public Construction Commission e-Procurement)

Try asking:

  • Taipei City government awards last month with vendor + amount
  • Total winning bid amount for a vendor across 11 years
2026-05-13v1.12.0

Legislature: county councilors + named votes

rows added

+1,453

datasets

2

  • ly-councilors: 603 rows of councilors from 22 cities/counties (Interior Ministry, OGDL v1)
  • ly-votes: 850 named-vote events in the 11th legislative session (extracted from session records)
  • Sourced from Interior Ministry + Legislative Yuan official data.ly.gov.tw

Try asking:

  • Taipei City councilors with party affiliation
  • Named-vote pattern on energy bills in the 11th session

Why this exists

Taiwan data is scattered across 100+ portals.
AI app builders don't have time to wire each one.

Twinkle Hub bundles them into a single MCP service: discover datasets, fetch rows, join across tables, pay — all on one endpoint. Phase one is making data.gov.tw excellent; more sources (industry associations, local governments, commercial data) come next.

01

One MCP endpoint

Every source, every domain, every tool under the same URL. Configure your client once.

02

Deterministic billing

Fixed price per tool — no token-based gambling. Prepaid wallet, hard cutoff, no surprise bills.

03

License compliance

Original license metadata is passed through end-to-end so downstream apps don't trip license terms.

04

MCP-native protocol

tools/list / tools/call are native endpoints, not OpenAI tool-calling translated. Claude / Cursor / OpenClaw / Hermes / Continue / Cline / your own agent loop — same wire format, no SDK, no adapter.

Marketplace preview

20 domains covering all of data.gov.tw

From real estate, healthcare and environment to foreign affairs and agriculture — all 52,960 datasets are auto-classified.

See the full marketplace

Real Estate & Land

realestate_land

Land, buildings, urban planning, land prices, building permits, property transactions, rents.

Typical question

「某地段近一年實價中位數」

See domain detail

Transport, Roads & Parking

transport

Highway, rail, air, marine, parking, traffic flow, accidents.

Typical question

「某路口近一年事故數」

See domain detail

Healthcare, Food & Drugs

health_food

Medical facilities, NHI, food safety, drug approvals, disease statistics.

Typical question

「住家附近健保藥局」

See domain detail

Why Twinkle Hub

Downloading 53,000 datasets is easy. Making them usable to AI is the hard part.

Making Taiwan open data usable to AI isn't a one-off catalog download. We handle the cleanup, conversion, classification, and querying — your AI just calls the API instead of crawling a hundred different portals.

52,960 / 19

53,000 datasets, sorted automatically

All 52,960 datasets on data.gov.tw, sorted into 19 categories. New ones land in the right place the next day — the index never ages out.

10+ formats

All the file formats, one shape

CSV, JSON, Excel, PDF, geo files — whatever the format, we read it and clean up the column names. Your AI doesn't choke just because one source is PDF and another is XLSX.

~50 ms

Ask a question, answer in 50 ms

Want your AI to query data on specific conditions? We have a query layer — about 50 ms across all 53,000 datasets. We don't dump a zip and tell you to process it yourself.

Daily refresh

Synced with the government every day

A daily cron pulls the latest catalog, picks up what's new or retired, runs classification. The data stays fresh — it doesn't go stale because we forgot to update.

Beyond the data

Two things we wired up alongside

OpenData is the main course. These two also matter for agents — we did them too.

32 TW utility tools

Your local model won't hallucinate Taiwanese addresses

Open-weights models are unreliable on Taiwan-specific tasks — wrong tax-ID checksums, made-up street numbers, garbled agency names. Twinkle Hub's 32 TW utility tools are deterministic functions — pass an address in, get a normalised address + district code + postcode out. No model guessing.

See all 32 tools

BYO MCP connectors

Notion / Sentry / n8n etc — install in your local client

These run in your local Claude Desktop / Cursor, tokens stay on your machine — twinkle-hub never collects, stores, or proxies them. We just ship the install guide + a one-click JSON snippet. You own the whole path.

See 6 connectors

Who it's for

Three roles, three paths

AI user

Paste a prompt into Claude — start querying Taiwan data

No code. Claude Desktop, Cursor, ChatGPT, OpenClaw, Continue, Cline — any MCP-aware client. Paste the prompt from the homepage and the AI sets itself up.

User docs

Developer

Bake the data into your AI app

curl / fastmcp Python / your own agent — call it any way you like, with transparent per-request billing you can audit.

Dev docs

Self-host / open-weights

OpenClaw, Hermes Agent, or your own agent loop

Full local / self-host stack: vLLM, SGLang, Ollama running your model; OpenClaw, Hermes Agent, or your own loop running the agent; Twinkle Hub feeding it Taiwan data. No commercial inference provider in the path. Our MCP endpoint is open at the protocol level — change one URL and you're gone.

See self-host docs

Ready to plug Taiwan data into your AI?

Free during alpha — no usage cap.

Get started