Getting started
What is NEM Explorer?
An AI interface to Australia’s National Electricity Market. Ask a question in plain English and get an answer built from real AEMO data, with the SQL, a chart and a table behind it. You don’t need to write SQL or code. There are two ways to use it:
- Web chat: a fast, low-cost model. Good for getting to know the data and for direct questions.
- MCP server: connect your own agent (Claude Code, Codex and others) for much deeper analysis.
Both get the same validated capabilities. Ask “how much money did Hornsdale BESS make on 1 April?” and the agent runs a pre-written, tested query that handles linear ramping between dispatch targets and counts FCAS income as well as energy.
How do I ask a question?
Type it into the chat box, or connect your agent to the MCP server. For example:
- What was the average spot price in Victoria last week?
- Compare wind and solar generation in South Australia over the past month.
- Show interconnector flows between New South Wales and Queensland for the last seven days.
What kinds of NEM questions can I ask?
Anything about the wholesale market and the data AEMO publishes on NEMweb, including:
- Spot prices, negative prices and price spikes by region
- Generation by fuel type, power station or DUID
- Interconnector flows and constraints
- FCAS prices, enablement and revenue
- Demand forecasts and pre-dispatch
- Bidding and rebidding behaviour
- Capacity factors and generation mix trends
Chat vs. agents
What does the web chat answer well?
One-shot questions. You ask something concrete, it picks the right tables, runs the query and writes the answer. It is fast and consistent for:
- Lookups, such as the NSW1 spot price on a given date
- Rankings, such as the top batteries by revenue in a quarter
- Standard aggregations, such as QLD fuel mix over the last 30 days
- Questions about specific DUIDs, regions and intervals
When should I connect an agent instead?
When the question needs several steps: inspecting tables, refining queries and following a trail. Connect an agent for:
- Open-ended exploration, like working out what is interesting about battery profitability
- Multi-step research, like finding every binding constraint and estimating lost revenue for affected DUIDs
- Finding patterns and anomalies, such as when a unit departs from its usual behaviour
- Counterfactuals, such as what might have happened without a constraint or a unit
- Compound questions with several parts
When the chat spots a question like this, it suggests the agent route rather than giving a shallow answer.
What does “Connect an agent” do?
It gives you the NEM Explorer MCP server URL and step-by-step setup for Claude, ChatGPT/Codex and Gemini Spark. Any other agent that supports remote MCP servers works too. Sign-in uses your existing NEM Explorer account. See the setup guide.
The agent gets tools to search entities, inspect tables, run validated capabilities and execute SQL. Large results come back as a 100-row preview plus a handle the agent can page through, so a big query never floods its context. The full list is in the docs.
In Claude you can also add it as a plugin from the brensch/scadaminer-ai-toolkit marketplace: in Claude Code, run /plugin install scadaminer --marketplace brensch/scadaminer-ai-toolkit. The plugin installs the same connector, so the tools stay current with the server.
Data & coverage
Where does the data come from?
AEMO’s NEMweb, the official public portal for National Electricity Market data: dispatch, SCADA, trading intervals, bids and offers, constraints, interconnector flows, FCAS, demand forecasts and more.
Which regions are covered?
All five NEM regions: New South Wales (NSW1), Victoria (VIC1), Queensland (QLD1), South Australia (SA1) and Tasmania (TAS1).
What time resolution is available?
It depends on the dataset. The finest is 4-second: frequency and per-unit MW measurements from AEMO’s frequency performance payments data (December 2024 onward), plus the discontinued causer pays archive (2011 to March 2017). Dispatch prices, generation and SCADA are 5-minute, pre-dispatch forecasts are 30-minute, and settlement, registration and PASA data run from daily up to longer horizons.
Is a “day” a calendar day or a trading day?
A calendar day in NEM time, which is fixed at UTC+10 all year with no daylight saving. So “yesterday” or “last Tuesday” means the NEM calendar date, not UTC or your local time zone.
How far back does the data go?
It depends on the dataset. Core dispatch and pricing goes back to 2009. Coverage is tracked for every table, and you’ll be told when a question falls outside it.
Accuracy & limits
How accurate are the answers?
The underlying data is as accurate as AEMO’s own publications. The layer that turns a question into SQL is tested against references such as the Open Electricity API and AER published figures.
As with any AI system, a question can still be misread, so check the query, filters and time range match what you meant. More on how answers are validated.
Is rooftop solar included?
Not as direct generation data by default. Rooftop and other behind-the-meter generation isn’t centrally metered by AEMO, so generation figures cover utility-scale units in the wholesale market. Where it could matter, the answer notes how rooftop solar was treated.
Why might generation totals differ from other sources?
Energy is estimated from dispatch data (TOTALCLEARED and INITIALMW) with a trapezoidal approximation across 5-minute intervals. Sources that use metered data, which is published later, can differ slightly.
Some smaller solar and wind farms report SCADA only intermittently or not at all, so for aggregate generation, dispatch data usually gives more complete coverage.
Is distribution network (DNSP) data included?
No. NEM Explorer covers wholesale market data only, not local usage, demand charges or DNSP tariffs. Get in touch if you need it.
Is retail pricing included?
No. Retail tariffs and demand charges aren’t available; the focus is the wholesale spot market and related AEMO data. Get in touch if retail pricing matters to you.
Pricing & credits
How much does it cost?
NEM Explorer runs on credits. Current packs and prices are on the Usage page under Buy credits.
How many credits does a question use?
It depends on the work involved. A simple price lookup might use around 50 credits; an analysis joining several datasets can use 200 or more. The cost is shown after each question.
Are credits shared between chat and agents?
Yes. One balance covers chat answers and the data queries your connected agent runs, so there’s nothing separate to top up for agent access.
Is there a limit on what a connected agent can do?
Queries have no cap beyond the scan limits on each one. CSV exports are charged on file size: 2 credits plus 1 credit per 5 MB.
Free access is for reasonable use. If an account’s use looks like bulk extraction or copying the service rather than analysis, we can withdraw its free access, and its queries and exports are then charged at the normal rates. For bulk data, contact SCADA Miner.
Is there a free allowance?
Plans can include a free monthly allowance that refreshes at the start of each month. It doesn’t roll over: unused free credits expire at month-end. Your allowance is on the Usage page.
Do credits expire?
Free monthly credits expire at the end of the month they were granted. Purchased credits last until their own expiry, which depends on your plan. Check the Usage page for your balance and expiry dates.
Exports & requests
Can I export data or get API access?
Yes, through a connected agent:
- CSV exports. The agent can save any query result as a private, expiring download, charged on file size (2 credits plus 1 credit per 5 MB).
- Live data links. Any query can become a link that re-runs it and serves the latest rows as JSON or CSV, ready to drop into your own code or spreadsheet.
- Dashboards. Saved dashboards refresh against live data and can be shared publicly.
For bulk extraction beyond that, contact SCADA Miner.
How do I request a feature?
Use the contact page, or email support@scadaminer.com.
Still curious? Just ask it.
The fastest way to find out what NEM Explorer can do is to put a question to it.