The price list reads like a cliff: a free key at $0, then the first paid plans at $99 a month, as the pricing page lists them, checked 2026-10-17. If your project sits in between, the honest answer is that no published plan fills the gap. The good news is that the gap is smaller than it looks, and there are two honest ways across it.
What the free key actually covers
The free tier's 100 requests a day buy more than they seem, because one request returns a whole sport board. The free key page works through what fits: a board refreshed every 15 minutes all day is 96 calls, inside the allowance, checked 2026-10-17. A prototype, a small dashboard, a research project on a daily rhythm: all of these live comfortably at $0.
The gap appears only when you need faster than every few minutes, or when you need drops pushed to you. Neither exists between free and $99; you either stretch the free allowance or step up.
Stretch the free side first
The free key page's own advice is the way to stretch: the since cursor keeps each call small, and a deliberate rhythm keeps the day inside the cap. Write the loop on the free key, prove the value, and only then let the price question in. Most projects that feel stuck in the gap are really at the prove-it stage.
Jump with a measured month
If the proof says you need the rate, the jump is $99, not $99 forever: buy one month, measure the loop against the allowance, and let the month's own numbers decide the renewal. The plan guide's order is the same: the loop picks the plan, and a month of logs is the loop made visible.
The demo covers the other gap: if the question is the transports, the three-day full demo with SSE and raw WebSocket is free on request, per the same pages.
A cliff is only a problem if you stand still in front of it. Stretch the free key while you prove it, then take one measured month. The gap was never meant to hold anyone.