Singapore's open data is a firehose — and it's free, even for commercial use.

Why this matters: a capability map, not an analysis

Before we build anything, it helps to know what raw material is on the table. This report maps the data that Singapore's government publishes for free, and explores what we could build on top of it.

Singapore's official data is unusually open. It is centralized on data.gov.sg and SingStat, machine-readable, and free for commercial use under the Singapore Open Data Licence. The practical upshot is that a single person or small company can pull the same numbers that inform government policy, at no cost.

This report is anchored to the threads we have discussed: the biscuits factory and food business, the OPC and business intelligence, property investment (now the Singapore angle after Japan and China), the Apple health dashboard, the kids and their education, and geospatial and market research. For each, the structure is the same: agency → what's free → how to access it → what we could do with it.

None of the "what we could do" ideas here are conclusions. They are possibilities, tagged as exploration, with the source each one would draw on.

The central platforms

Everything else in this report sits on two foundations: data.gov.sg, the central catalog, and SingStat Table Builder, the deep statistical store.

data.gov.sg — the anchor

You can browse and download datasets as CSV, XLSX, GeoJSON and JSON, or hit REST APIs (Real-Time and Collection/Dataset). Public endpoints work without a key for testing; a free sign-up unlocks production keys with higher rate limits. Exceeding a limit returns HTTP 429. The developer guide explains the API overview, Real-Time APIs, how to request an API key, and the rate limits.

A working example: the collection metadata for Resale Flat Prices (collection 189) returns real HDB transaction data from 1990 to 2025, updated monthly. The licence is the Singapore Open Data Licence v1.0 — free forever, commercial use allowed, attribution required.

SingStat Table Builder

SingStat Table Builder covers GDP, CPI, trade, employment, demographics and prices. The developer API returns JSON (100 calls/min per IP), and tables download as CSV or Excel. Use the find-APIs tool to locate a table's endpoint. Notably, SingStat launched an MCP (Model Context Protocol) in 2026 that lets AI agents query the tables securely. Key tables referenced later: Manufacturing production index (M355352), F&B services index (M602161), business expectations (M250142), residents by age group (M810011), by planning region (M810771), and by dwelling type (M810381).

How they fit together

data.gov.sg is the central catalog and machine-readable mirror. SingStat is the deep, customizable store with an API and an MCP. Real-time and transactional data (LTA, OneMap, URA, MAS) lives on the agencies' own portals, each with its own sign-up. The combination means most needs can be met with free, official, machine-readable data.

The biscuits factory & food business

For the cousin's factory and food industry work, the free official data splits into production, demand, licensing, and trade.

For context, SSIC 10711 maps to "Manufacture of biscuits". SFA is the licensing authority for food manufacturers. There is no dedicated Singapore Customs API, so trade figures come via SingStat/data.gov.sg plus WITS and Trade Map.

Exploration ideas: an output-and-demand dashboard for the food segment, a food-trade picture for biscuits and snacks, a competitor scan through ACRA, and a CPI-based cost/pricing view. Each is possible with free data, but none is a conclusion yet.

Property investment

After the Japan and China/Shenzhen threads, this is the Singapore angle. The free property data here is unusually rich, especially HDB resale.

URA REALIS holds deep history at project level but is a paid subscription; the free 60-month window on the PMI e-Service is generous. ABSD for foreigners and companies buying residential property can reach 65%, which changes the maths entirely, so it is a key input.

Exploration ideas: a Singapore property dashboard (HDB + URA + price index), a return analysis with ABSD and SORA baked in, and a cross-country comparison against the Japan/China/Shenzhen figures we explored. All possible, none final.

The health dashboard

For the Apple Watch and dual-profile health dashboard, the free official data is environmental and population-level — not individual.

What is not open: individual or clinical health data (HealthHub) is private, and life expectancy and mortality are published as tables and PDFs on the MOH site, not as a public JSON API. We should not claim either exists as open data.

Exploration ideas: fold real-time air quality and temperature into the family health dashboard, and use MOH attendance data as context. All exploratory.

The kids & education

For Isaac (9, P3) and Matt (13, Sec 1 planning) and the tutoring work, the free data covers schools, demographics and nature.

What is not open: MOE does not publish PSLE scores or school rankings as datasets; it only offers an interactive SchoolFinder. We should not claim PSLE data exists as open data.

Exploration ideas: a school comparison tool from the MOE directory, a school-age population map from SingStat, and nature/geospatial learning projects from NParks. All possible, none final.

Transport, environment, geospatial

The broader geospatial and transport layer that connects many of the other threads.

LTA DataMall requires a free Account Key for dynamic APIs; OneMap now requires token registration for its Search API. COE prices, ridership and ERP rates are especially useful for analysis, and real-time bus and traffic feeds suit transport apps.

The big picture: what we could build

A consolidated possibility map, grouped by interest. Every row is an exploration idea ("could do"), not a conclusion, and each is tagged with its data source.

The through-line: nearly every interest we have maps to free, official, machine-readable data. The constraint is never cost — it is time and effort to pull, join and interpret the datasets.

The SingStat MCP — an AI-agent hook

Directly relevant to us: the SingStat MCP lets AI agents query official Singapore data.

Launched in 2026, the SingStat MCP lets AI agents query official Singapore statistics securely, bridging to SingStat Table Builder's 2,500+ tables. For an agent like Hermes or Zeus, this means official Singapore figures could be pulled directly into dashboards, briefings and market research without manual scraping or tabular downloads.

There are also third-party wrappers (for example, the unofficial mcp-singstat-sg). These are unofficial. The official SingStat MCP is the one to use; treat third-party wrappers as untrusted until verified.

Exploration idea: wire the official SingStat MCP into our agent workflows to fetch live official statistics on demand. Possible, not yet set up.

Licensing & rules

So we stay on the right side of the terms.

• Singapore Open Data Licence v1.0: free forever, for personal or commercial use. Attribution required (dataset name, access date, source, licence link). No endorsement implied. Excludes personal data, third-party rights and trademarks. Datasets are provided as-is.
• Respect rate limits (HTTP 429 on excess): data.gov.sg Real-Time 6/12/30, Datastore 4/8/20, Downloads 2/4/10 per 10s; SingStat 100 calls/min/IP; LTA and OneMap need an account or token.
• Personal health data (HealthHub) is not open. Never assume individual-level data exists.
• Do not claim data that is not published: PSLE scores or school rankings, individual health, or a life-expectancy JSON API.

Two practical notes. First, a few government portals are JavaScript single-page apps that may not render server-side; the official documentation pages are authoritative. Second, every source in this report was checked with an HTTP request that returned 200, so the links are live.

Sources

All verified live at build time. A few government portals are JS single-page apps that may not render server-side; official docs pages are authoritative.