What Is a Taiwan Stock Market API? A Complete Introduction

What a Taiwan stock market API is

A Taiwan stock market API turns the public filings of the Taiwan Stock Exchange (TWSE), the Taipei Exchange (TPEx) and the Market Observation Post System (MOPS) into an interface your code can call directly. Instead of scraping pages, unpacking CSVs and converting ROC-calendar dates and legacy encodings yourself, you pull prices, monthly revenue, institutional investor flows and corporate events straight into a backtest, a dashboard or an AI agent workflow.

Worth saying plainly: an API of this kind does not produce the data. The data is already published on official sites, and it is free. What the API does is collapse the same public filings — scattered across different systems, formats and release times — into one set of fields and one way to query them, so that "give me every daily bar for 2330 since 2015" is a single request rather than a scraping project.

Who publishes Taiwan market data

The thing that most often trips people up early is assuming there is one source. Taiwan's public market data is split across four bodies, each with its own site, field vocabulary and release cadence:

• TWSE (Taiwan Stock Exchange) — listed-market daily prices, trading statistics, and institutional investor net buying.
• TPEx (Taipei Exchange) — the same for the OTC and emerging markets, on its own domain and under different field names.
• MOPS (Market Observation Post System) — company filings: monthly revenue, the three financial statements, material announcements and company profiles.
• TAIFEX (Taiwan Futures Exchange) — futures and options.

Every Taiwan data product, this one included, ultimately sources from those four. The difference is never whether the data exists — it is what happens after it lands: whether fields are normalized, whether listed and OTC line up, whether history is accumulated, and whether each row can be traced back to where it came from.

What you can actually pull

"What is there besides prices?" is the most common follow-up. With TW Market Data it falls into four groups:

• Daily prices — TWSE from 2004 and TPEx from 1994, keeping the full price history of names that have stopped trading.
• Monthly revenue — the MOPS monthly revenue filings, from 2010-01 across 2,096 companies, with YoY and MoM. How much of that history you can query depends on your plan.
• Institutional investor flows — net buying and selling by foreign investors, investment trusts and dealers, as its own dataset, queryable by symbol and date.
• The three financial statements, dividends and corporate events — also from MOPS filings.

Every dataset states its own coverage window and known gaps on its own page; there is no blanket claim that everything is covered. The same data suits a program or an AI agent: responses use a consistent structure carrying source_role and lineage, openapi.json plus llms.txt let an agent discover the endpoints on its own, and developers call the REST API with an API key.

How to read a Taiwan dataset

Whatever dataset you are handed, three properties decide whether it can carry a decision:

• Coverage — the first and last date, and how many instruments. Without a stated window you cannot tell whether a blank stretch means the market did not trade or the provider does not have it.
• Lineage — which official endpoint a row came from, and whether it is a raw or a derived value. Every TW Market Data response carries source_role and its source markers.
• Data gaps — whether holes are labelled or quietly filled with interpolated values. This is the one most often skipped and the one that hurts most: a source that fills gaps silently makes a backtest look better than reality. We preserve disclosed gaps rather than filling them.

A fourth question is worth asking: are delisted names still there? Most price datasets drop a stock entirely once it stops trading, so your historical universe contains only today's survivors and backtest returns come out inflated — that is survivorship bias. TWSE daily prices currently start in 2004 and cover 1,682 listed names across roughly 5.2 million rows, including 311 names that have stopped trading but keep their full price history (262 of them with an official delisting date). TPEx daily prices start in 1994, and monthly revenue starts at 2010-01 across 2,096 companies.

Three questions to ask before you pick one

Three questions are worth asking of any Taiwan stock data API: does the data come first-hand from the official source, can every row be traced back to where it came from, and are the coverage window and the known gaps disclosed honestly? Asking those three separates most options on its own.

What the free tier covers

Is there a free Taiwan stock market API? Yes, in two senses. The official TWSE and TPEx OpenAPIs are themselves free first-hand sources and need no registration. On our side, the free plan gives you 500 requests per month, 1 API key and 60 requests per minute across the core datasets (the three financial statements are excluded), for non-commercial use. Five symbols (2330, 2317, 2454, 0050 and 2603) also work with no key at all, so you can get an integration running end to end before deciding whether to upgrade. Per-plan quotas and queryable depth are listed on the pricing page.

How this differs from the official TWSE OpenAPI

The official OpenAPI is free, authoritative and the first-hand source for all of this data — if you are happy to do the plumbing, it is enough on its own, and that is worth stating plainly rather than talking it down. The practical differences cluster in three places: official endpoints are separate, each with its own fields and format, so joining across datasets is your job; most open endpoints serve the current period, so historical depth is something you accumulate yourself over time; and request quotas and formats are whatever the official announcements say at the time.

TW Market Data does not replace the official source — it takes the same official data, puts it behind one consistent schema, and adds historical depth and source lineage so a system or an AI agent can answer a question in a single query. If you only need the current snapshot and are willing to write the parsing and accumulation yourself, calling the official endpoints directly is a perfectly sensible choice.