Building an AI agent? Start with /llms.txt for the full site index.
Overview
How far each data family reaches, how often it updates, and how to read the data_gaps signal.
The left-hand navigation lists every dataset page, and Quick Start walks through building a first request.
For agents and internal tools that need to read the documentation programmatically, TW Market Data publishes /openapi.json, /llms.txt and /llms-full.txt as machine-readable entry points, and an MCP server at https://mcp.twmarketdata.com/mcp.
Coverage is the range a dataset actually spans — across tickers, dates, reporting periods or event types. It moves as sources publish and as datasets are productised, so no dataset should be assumed complete.
Treat coverage as an observable state rather than a constant. The API response and the dataset's own documentation are what govern, not an assumption made once at integration time.
/sell: coverage is still being filled in — read it alongside data_gaps./preview/deferred status on the docs page rather than assuming the dataset is fully live.Freshness is the last time a given row was updated and available. Cadences differ by dataset: daily prices, monthly revenue, quarterly statements and event data are not published on the same rhythm.
When a query spans several datasets, check freshness together with each dataset's own time columns. Data that arrived on different cadences is not data from the same moment.
data_gaps marks a hole in the data or a limitation at the source. It is a signal worth keeping in a research pipeline, not a zero and not something to fill in automatically.
Carry data_gaps, source_role and lineage through the pipeline together, so a result can be audited and traced back later.
These are the status terms the platform uses, so that availability and its limits can be told apart inside your own systems.
Read the freshness, source_role, lineage and data_gaps fields on the response, then check them against the endpoint's own documentation and its parameter limits.
The dataset index (/datasets) and the individual dataset pages — /datasets/twse-daily-price, /datasets/institutional-flow — carry the same facts in longer form.
For a specific industry, a dataset that is not listed, or an enterprise deployment, get in touch.