One credit is 1,000 tokens of AI work. A full clinical synthesis of one resident — the med list, the flags, the interactions, the narrative — averages 6,121 tokens, or about 6 credits. The smallest paid plan’s 500 credits is roughly 81 full resident reviews a month.
Going over does not stop your clinical work. We warn you at 80% and again when you pass the allowance, and we keep running reviews. Declining to summarize a resident carrying a critical interaction in order to save a few cents is not a trade we are willing to make. The only hard stop on that pool is a runaway guard at 10× your allowance, which exists to catch a loop, not a busy month.
The AI assistant draws on its own separate pool, so a long conversation can never eat into the credits your clinical reviews need. That pool does pause at its limit — the assistant stops answering until the month turns over, while reviews carry on. We would rather say so here than have it surprise you.