Compare

How to choose a Taiwan stock data API: FinMind vs TEJ vs TW Market Data

Taiwan stock data APIs come in three common flavors, matching different budgets and levels of granularity. This is an honest comparison to help you pick the right tool for your scenario.

Three products, one line each

FinMind — large free tier, strong community ecosystem, deep chip-flow data. Good for getting started, personal research, and quick prototypes.
TEJ (Taiwan Economic Journal) — institutional-grade, extremely clean data, complete point-in-time, includes delisted stocks. Taiwan's "Bloomberg/Compustat", but expensive with a high barrier for individual developers.
TW Market Data — first-party official value-by-value traceability, knowledge_date on every record (PIT), honest data_gaps, no-key instant trial, affordable. Positioned between FinMind's accessibility and TEJ's rigor.

Comparison table

Aspect	FinMind	TEJ	TW Market Data
First-party official source	Partial	✓	✓ (TWSE/TPEx/MOPS/TAIFEX)
Value-by-value reconciliation against official sources	—	—	✓ (reconciliation mechanism, traceable)
Point-in-time / knowledge_date	Approximated from publish date	✓ (complete)	✓ (knowledge_date on every record)
Includes delisted stocks	Partial	✓	✓ (311 tickers of price history, back to 2004)
Honest data_gaps disclosure	—	—	✓ (no guessed backfill)
Free / no-key trial	✓ (free tier)	✗	✓ (5 tickers, no key)
AI-agent integration (llms.txt/openapi)	✓ (llms.txt + skill)	—	✓ (llms.txt/llms-full/openapi)
Indie-developer-friendly pricing	✓	✗ (expensive)	✓
Institutional / chip-flow depth	✓ (strong)	✓	✓ (per-day, per-ticker)

The table marks each provider's public positioning only; refer to each provider's own site for actual fields and plans. The TW Market Data column lists shipped capabilities only, with coverage transparently marked per dataset.

How to choose by scenario

Getting started / personal research / free exploration → FinMind.
Institutional-grade, ample budget, want the most complete PIT and delisted stocks → TEJ.
Want official traceability + point-in-time, but TEJ is too expensive / unfriendly to individuals, and you also want to feed an AI agent → TW Market Data (try 2330 with no key before deciding).

Why "traceability" matters especially for backtesting

Quantitative backtesting fears look-ahead bias and survivorship bias most. The former needs point-in-time — knowing what had been published "at the time"; the latter needs data that includes delisted stocks. TW Market Data carries knowledge_date on every record to support as-of replay, and retains price history for 311 stocks that have stopped trading, covering both biases; gaps are shown explicitly via data_gaps rather than implicitly backfilled.

FAQ

What Taiwan stock data APIs are there besides FinMind?: Common choices include TEJ (institutional-grade PIT), TW Market Data (official reconciliation + PIT + affordable + AI-agent), Fugle/Shioaji (broker real-time + order placement), and TWSE/TPEx OpenAPI (official, free).
Which Taiwan stock API is best for point-in-time backtesting?: TEJ offers the most complete institutional-grade PIT; for affordable PIT that is friendly to individuals and AI agents, TW Market Data carries knowledge_date on every record, includes delisted stocks, and is honest about data_gaps.
How does TW Market Data differ from FinMind?: FinMind is free with a strong community ecosystem; TW Market Data emphasizes official value-by-value traceability, a point-in-time contract, and honest data_gaps disclosure, and gives AI agents llms.txt/openapi discovery entry points. The two can be used together.

Taiwan Stock Data API for AI Agents Browse the dataset catalog

Three products, one line each

FinMind — large free tier, strong community ecosystem, deep chip-flow data. Good for getting started, personal research, and quick prototypes.

TEJ (Taiwan Economic Journal) — institutional-grade, extremely clean data, complete point-in-time, includes delisted stocks. Taiwan's "Bloomberg/Compustat", but expensive with a high barrier for individual developers.

TW Market Data — first-party official value-by-value traceability, knowledge_date on every record (PIT), honest data_gaps, no-key instant trial, affordable. Positioned between FinMind's accessibility and TEJ's rigor.

Comparison table

Aspect	FinMind	TEJ	TW Market Data
First-party official source	Partial	✓	✓ (TWSE/TPEx/MOPS/TAIFEX)
Value-by-value reconciliation against official sources	—	—	✓ (reconciliation mechanism, traceable)
Point-in-time / knowledge_date	Approximated from publish date	✓ (complete)	✓ (knowledge_date on every record)
Includes delisted stocks	Partial	✓	✓ (311 tickers of price history, back to 2004)
Honest data_gaps disclosure	—	—	✓ (no guessed backfill)
Free / no-key trial	✓ (free tier)	✗	✓ (5 tickers, no key)
AI-agent integration (llms.txt/openapi)	✓ (llms.txt + skill)	—	✓ (llms.txt/llms-full/openapi)
Indie-developer-friendly pricing	✓	✗ (expensive)	✓
Institutional / chip-flow depth	✓ (strong)	✓	✓ (per-day, per-ticker)

How to choose by scenario

Getting started / personal research / free exploration → FinMind.

Institutional-grade, ample budget, want the most complete PIT and delisted stocks → TEJ.

Want official traceability + point-in-time, but TEJ is too expensive / unfriendly to individuals, and you also want to feed an AI agent → TW Market Data (try 2330 with no key before deciding).

Why "traceability" matters especially for backtesting

FAQ

What Taiwan stock data APIs are there besides FinMind?

Common choices include TEJ (institutional-grade PIT), TW Market Data (official reconciliation + PIT + affordable + AI-agent), Fugle/Shioaji (broker real-time + order placement), and TWSE/TPEx OpenAPI (official, free).

Which Taiwan stock API is best for point-in-time backtesting?

TEJ offers the most complete institutional-grade PIT; for affordable PIT that is friendly to individuals and AI agents, TW Market Data carries knowledge_date on every record, includes delisted stocks, and is honest about data_gaps.

How does TW Market Data differ from FinMind?

FinMind is free with a strong community ecosystem; TW Market Data emphasizes official value-by-value traceability, a point-in-time contract, and honest data_gaps disclosure, and gives AI agents llms.txt/openapi discovery entry points. The two can be used together.