Running a portfolio of solar and wind farms means living inside ISO market data: day-ahead locational marginal prices, nodal congestion, hourly profiles that shift with weather and demand. The data is public and plentiful — the problem is that it arrives as raw time series, and somebody still has to turn it into a decision.
The gap between available and usable
Every ISO publishes prices. Almost none of that publication is shaped for the question an operator actually asks: should I bid this asset into tomorrow's peak, and what does the downside look like if congestion moves?
Answering that means joining a settlement node to an asset, pulling a trailing window, cleaning intervals, computing percentiles, and reading the shape of the distribution — not just its mean. Done by hand, it is an afternoon. Done per asset, per market, per day, it never gets done at all.
Ask the agent, not the spreadsheet
On the Cyberia Agentic Platform, an Energy ISO agent works your assets directly. Ask about a wind farm or a solar site and the agent pulls the relevant pricing node and returns a complete analytical picture:
- Price preview — day-ahead LMP over the trailing week, with peaks flagged and hoverable values
- Summary statistics — mean and median $/MWh, sample counts and fetch windows stated explicitly
- Day-by-hour heatmaps — average price by day and hour-ending, so peak windows and negative-pricing risk are visible at a glance
- Distribution and profile analysis — percentile bands, histograms and cumulative curves that expose volatility, not just averages
- Top and bottom intervals — the 25 best and worst settlement periods, ready for bid strategy review
Reading a report in sixty seconds
The layout is deliberate. Mean and median sit side by side because the spread between them is the first signal of skew. The heatmap comes before the histogram because operators think in hours before they think in distributions. And the worst 25 intervals are always shown next to the best 25, because negative pricing exposure is the risk that actually damages a season.
Asset intelligence, on demand
The same workspace answers research questions about the assets themselves. Ask about a solar project and the agent compiles a structured brief — location, capacity, footprint, operator and status, economic and environmental impact — sourced, organised and rendered as a document instead of a pile of search results.
That brief is the artefact a development team needs when screening a new market: enough context to decide whether to keep looking, produced in the time it used to take to open the first tab.
From raw text to real action
This is the shift the Agentic Platform is built around: agents don't retrieve text, they produce insight. A trader reviewing tomorrow's bids, an asset manager screening a new market, an analyst writing a site report — all of them get decision-ready output in minutes, with the underlying series, percentiles and sources attached.
Three properties make that output trustworthy:
- Explicit windows — every statistic states the interval it was computed over, so nobody compares a seven-day mean to a thirty-day one by accident.
- Attached provenance — the node, the fetch time and the record count travel with the number.
- Consistent shape — the same report structure for every asset, so portfolio comparisons are valid without re-normalisation.
Renewable portfolios are data-rich and attention-poor. Agents close that gap by doing the analysis, not just the lookup.
Composing with the rest of the platform
Because the platform wires agents to data products, forecasting models and memory as first-class tools, these energy workflows compose with everything else — alerting hooks, scheduled runs, and downstream reporting — without bespoke integration work.
A scheduled run can post the morning bid brief before the desk opens. A forecasting model from the Data platform can sit next to the ISO tool in the same agent. A memory of last quarter's congestion events can shape how this quarter's anomalies are flagged. None of that requires a new pipeline — it requires attaching another tool to an agent that already knows the assets.
Where to start
Pick one market and one asset class. Connect the settlement nodes, define the report once, and schedule it. The value shows up on the first morning the desk reads insight instead of assembling it.