---
title: "Usage allowance"
description: "How Tunang's usage allowance is counted, what the meter states mean, and what happens when it runs out."
---

> Documentation Index
> Fetch the complete documentation index at: https://docs.divelogic.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Usage allowance

import { SettingsTab } from "@/components/react/AccountScreens";

Every answer Tunang gives is counted against your account's allowance, wherever
you asked for it. The authoritative view is Settings → Subscription, under
**Current plan**.

## The states

| Meter state | What the app says | What it means |
|---|---|---|
| Within allowance | "AI usage is within your allowance." | Nothing is limited. |
| Running low | "Most of your AI allowance for this period is used." | You have crossed 80 %. Requests still go through. |
| Used, refreshing | "Your AI allowance for this period is used. Assistant requests resume when it refreshes." | Requests are declined until the period turns over. |
| Used, no refresh | "Your AI allowance is used. A higher plan restores assistant access." | A one-time grant. Only a plan change restores access. |

Below the state, the tab names the period: "Refreshes on *date*.", "Your plan
ends on *date*.", or "This allowance does not refresh." A monthly allowance
refreshes on your billing anniversary.

## What is counted

Not messages, and not words, but how much work each answer takes, which is why
two questions of the same length can cost different amounts.

- **A long chat costs more per answer than a short one.** The whole conversation
  is read again for every message, so a new chat for a new topic is cheaper.
- **A calculation is not free.** A question containing three variations costs
  roughly three answers.
- **A refused answer costs nothing.** Once the allowance is spent, requests are
  declined before any work is done.

The app shows a state rather than a figure; there is no counter in the
interface.

## When it runs out

**Crossing 80 %.** A card headed **AI allowance** appears in the thread with the
meter, "Most of the AI allowance for this period is used." and "Assistant
requests continue until the allowance is used." It is announced once per period,
and a notification titled *allowance running low* arrives with it.

**Reaching the limit.** The next question is refused, and the reply is the same
card exhausted: "The AI allowance for this period is used." and "Assistant
requests are declined until the allowance renews or the plan changes." Both
carry an **Upgrade plan** button. It can also arrive after an answer whose own
cost finished the allowance off; that answer stands. The post-dive debrief has
no chat to put a card in, so its refusal is written into the debrief text.

## Next

- [Subscription and billing](/account/subscription) covers changing tier.
- [Notifications](/account/notifications) puts the two allowance notifications
  in context.

Source: https://docs.divelogic.ai/assistants/allowances/index.mdx