> ## Documentation Index
> Fetch the complete documentation index at: https://docs.make-pretty.com/llms.txt
> Use this file to discover all available pages before exploring further.
# Usage and limits
> The monthly AI budget in euros, where to watch it, what consumes it, and how to get more done for less.
Pretty meters what your chats cost in AI usage and compares it against a monthly budget. When
the budget is gone, new requests stop until the month rolls over or an admin raises it.
## Where to watch it
Open the settings drawer and go to **Usage**. The panel shows one line and a bar. Spend comes
first, then the budget it counts against:
```text theme={null}
€24.60 of €200 this month
```
Underneath, either *"Resets on the 1st of each month"* or, once you are out,
*"Monthly limit reached — new requests are paused"*. The bar turns amber past 80% and red at
100%, and a **Monthly limit reached** banner appears in the pane.
The figure is **your own** spend for the calendar month, measured against the per-person budget
your workspace sets. A colleague's heavy week does not eat into your allowance, and yours does
not eat into theirs.
## Hitting the limit
Send a message with nothing left and the chat answers with:
> You've reached your monthly AI limit. Please contact an administrator to raise it.
Nothing is lost. The chat, its history and its attachments are all still there. It will not run
another turn.
Two ways forward:
* **Wait for the reset.** The counter starts again on the 1st.
* **Ask an admin to raise the budget.** There is no budget field in workspace settings, so an
admin raises it by writing to [support@make-pretty.com](mailto:support@make-pretty.com) with
your workspace name and the figure you want.
Workspaces running on their own provider key are billed by that provider instead, and this cap
does not apply to them at all. See [Model providers](/admin/model-providers).
## What consumes the budget
Cost tracks how much work the model does, so the expensive turns are the ones with a lot to read
or a lot to produce.
| Heavier | Why |
| --------------------- | ---------------------------------------------------------------------------------------------------------------------------- |
| Long chats | Every turn re-reads the conversation so far. A chat that has run for an hour carries an hour of history into its next reply. |
| Research turns | **Researching** fetches live pages and feeds the prose back into the conversation. |
| Vision review | **Reviewing design** and **Snapshot** send rendered images of your slides to the model. Images are not cheap to look at. |
| Big attachments | A 60-page PDF is read in full before the first slide is drafted. |
| Whole-deck operations | "Restyle every slide" is one instruction and thirty slides of work. |
| Lighter | Why |
| ----------------------- | --------------------------------------------------------------------------- |
| Short, specific edits | "Make this heading two lines" reads one shape and writes one shape. |
| A fresh chat per topic | Nothing irrelevant to carry. |
| Lower capability models | Less reasoning before the model acts. See [Choosing a model](/chat/models). |
## Spending less
The single biggest saving. A chat that built a market-sizing slide is dead weight when you
move on to the appendix, and every subsequent turn pays to carry it. **New Chat** costs
nothing.
Extract the ten relevant pages rather than attaching the whole report. If you only need one
table, paste the table.
If three slides out of twenty are wrong, ask for those three. Re-running the whole deck
redoes the seventeen that were already fine, and may change them.
A vague brief costs a draft, a correction and a re-draft. Say the audience, the message and
the constraint up front. See [Prompting](/chat/prompting).
High capability earns its cost on a dense slide built from a source document. It is wasted
on "shorten this bullet".
Check **Usage** before a big build, not after. Assembling a full deck from a brief is the most
expensive thing Pretty does, and it is unpleasant to have it stop halfway.