Most of an agent’s reads on a quiet day return what the last read returned. The exceptions cluster around scheduled events. A data release, a rate decision, an earnings date. The times are public. An agent that uses the schedule as its clock spends its effort where the readings actually change.
Start with the schedule
get_economic_calendar returns the scheduled US releases with their dates and times. These are the moments that reprice volatility. At the start of the week, have the agent list them in order, with the time of each in your own time zone. That list is the plan for the week’s reads. Which releases tend to matter for the index is covered in ranking the economic calendar for SPY.
For a single stock, the date that matters is earnings. get_earnings_dates passes through a public exchange calendar. It is fetched, and NoVo did not measure it. If the upstream fails, the tool says so by name. An agent should report that as a failed lookup and never as no earnings scheduled.
The read before
Ahead of a release, the agent takes a snapshot. Where volatility ranks, from get_volatility_record. Where the walls and spot are, from get_dealer_levels. On the free tier the expected move is gated, and the report should say withheld. A paid key opens get_equity_map, which carries it. What that figure means is in the expected move.
Each line gets its age. A snapshot taken an hour before the release and one taken a day before are different pictures.
The read after
After the release, the agent reads the same things again and sets the two snapshots side by side. Did a wall move. Is spot on the other side of one. Did the volatility rank shift. This is a comparison of two stored readings, and it needs the first one kept. The setup for that is in an agent that reports only what changed.
Allow for delay. The free levels are delayed by design, so a read taken moments after a release may still describe the market before it. The age on the reading tells you which side of the event it sits on. Have the agent state it.
What it must not do
It must not guess the number. A model asked about an upcoming release will happily offer what it expects. Nothing in the tools supplies that. It must not say how the market will react either. The readings describe positioning and volatility before and after. They do not rank outcomes.
It should also not treat every calendar line as equal or decide for itself which ones matter. Give it your own short list of the releases you care about.
On days with nothing scheduled
The agent says so. No releases today is a useful line. It tells you the day’s moves will have to come from somewhere else, and that a routine read at the usual time is enough.
Where NoVo fits
The calendar, the volatility record and the dealer levels are free tools on the MCP & API. Trader draws the same map with the expected move on it. How scheduled events and volatility interact is in event volatility and scheduled catalysts.