logo
Forecasting Agent

/forecast from a single prompt

Use the forecasting agent in Claude or ChatGPT. Describe what you want to forecast, and get back a chart, a summary and the reasoning behind it.

Ask in plain language

Type this in Claude or ChatGPT with the Chronulus plugin installed:

/forecast monthly foot traffic across the Brooklyn Bridge over the next 12 months.
Researching, forecasting and plotting…

The agent sets up the use case, creates the forecast and returns the result shown here.

What happens after you send it

  1. 01

    The assistant researches the question

    After your prompt, the assistant (Claude, for example) looks for the latest data and context. Where it makes sense, it also finds the right scale for the forecast.

  2. 02

    The Chronulus agent makes the forecast

    The assistant passes its research to the Chronulus agent, which returns the forecast at the frequency and horizon you asked for, along with an explanation of the factors it considered.

  3. 03

    The assistant scales and plots it

    Finally, the assistant applies the scale to the forecast and plots it, labelling the periods of interest, the points in time marked on the chart.

Forecast of monthly pedestrian traffic across the Brooklyn Bridge, with a summary and the agent's explanation

A forecast and its reasoning

Every forecast comes with a written explanation of the seasonality, events and constraints behind the series, so you can check the logic before you rely on the numbers.

Real units, on request

Forecasts are returned on a 0 to 1 scale. Give the agent a minimum and maximum, or let it assume a range, and the series is rescaled into pedestrians, units or dollars.

Past, present and future

Forecast ahead, estimate a historical trend where no data was recorded, or nowcast what is happening now before the numbers are published.

No history required

The agent forecasts from a description of the situation, so it works for new products, new locations and anything without a data feed.

Frequencies and forecast length. Choose the frequency that fits your question, up to the longest horizon for each.

168 hours

Hourly · up to 1 week

365 days

Daily · up to 1 year

260 weeks

Weekly · up to 5 years

120 months

Monthly · up to 10 years

Use Cases

Anywhere you need a number over time and have no history to model it from. Pick an industry to see examples, and copy any of them as a prompt to try.

Investing & Markets

Size demand for new products and platforms when there is no track record to extrapolate.