# Tokens and usage

> What costs tokens, what does not, how the monthly pool and daily cap work, and what to do when you run low.

Tokens are how Vaulter meters AI work. This page explains exactly what spends them, what never does, how the monthly pool and daily limit interact, and what your options are when you run low. For plan sizes and prices, see [Plans and pricing](/docs/account/plans).

## What spends tokens

Tokens are spent when AI does work on your app:

- **Chat requests.** Every message you send the agent that reads or changes your app.
- **Edit Mode saves that need AI.** Deterministic style tweaks (colors, spacing, sizes) are applied directly, but edits that need the agent to rewrite code go through the same AI pipeline as chat.
- **The initial build** when a project is first generated from your prompt or a template change.
- **Plan mode**, which always runs on the most capable model.

How much a request spends depends on two things: how much of your app the agent has to read and rewrite, and [which model runs it](/docs/building/choosing-a-model). A copy tweak on Vaulter Mini might cost a few thousand tokens. A multi-screen feature on Vaulter Max can cost hundreds of thousands.

## What never spends tokens

Just as important is the list of things that are free, no matter how often you do them:

- **Auto-fix.** When Vaulter [repairs a broken build or a runtime error](/docs/building/auto-fix) on its own, that repair work is on us. It never comes out of your allowance.
- **The Undo button.** [Undoing the agent's last change](/docs/your-app/undo-and-versions) is a mechanical restore, not an AI operation.
- **Previewing.** The preview rebuilds after every change and that is never metered.
- **Publishing and rollback.** Publishing has its own daily count per plan, but it does not spend tokens, and neither does rolling back.
- **Browsing your data, editing code by hand, changing settings.** Only AI work is metered.

The corollary: if you can make a change yourself faster than describing it, the manual route is free. Fixing a typo directly in the [code editor](/docs/your-app/code-editor) costs nothing; asking the agent to fix it costs a (small) request.

## The pool, the period, and the daily limit

Your workspace has one token pool per billing period.

- **The monthly allowance** is set by your plan and, on Starter and Pro, by the ladder step you chose. On Teams the pool is shared: the seat size multiplied by the number of editor seats, spent by everyone in the workspace together.
- **The period follows your billing date**, not the calendar month. If your subscription renews on the 12th, your tokens reset on the 12th. On the free plan the period is the calendar month.
- **Unused monthly tokens do not carry forward** into the next period. Only purchased top-ups persist.

On top of the monthly pool sits a **daily limit**, so one runaway day cannot drain the whole month:

| Plan | Daily limit |
| --- | --- |
| Free | 300,000 tokens |
| Starter | 20 percent of your chosen monthly allowance |
| Pro | 20 percent of your chosen monthly allowance |
| Teams | None |

When you hit the daily limit, building with AI pauses until the next day even though monthly tokens remain. The tracker in the build workspace and the messenger both tell you when this is the case, and everything except AI requests keeps working in the meantime.

## Top-ups

Paid plans can buy **top-up tokens** at any time, in sizes from 1 million to 200 million.

Top-ups behave differently from the monthly allowance, deliberately:

- **They never expire.** A top-up balance persists across periods until it is spent.
- **They are spent last.** Your monthly allowance is always consumed first, so a top-up is pure extra runway.
- **They are not subject to the daily limit.** Once the monthly allowance is exhausted, or the daily limit is reached, top-up tokens keep you building.

Buy them from **Settings → Tokens** or from the prompt that appears when you run low.

## Watching your usage

- **In the build workspace**, the tracker pill in the header shows the pool and how much of it is used, live.
- **In Settings → Tokens**, you get the full picture: monthly usage, the daily limit, your top-up balance, and when the period resets.
- **On Teams**, admins additionally get per member usage, per member budgets, and a circuit breaker that stops any single member from burning more than a set share of the remaining pool in a short window.

## Making tokens go further

1. **Leave the model picker on Auto**, or drop to Vaulter Mini for small, well-defined changes.
2. **Be specific.** One precise request lands in one attempt; a vague one costs two or three rounds. The [prompting guide](/docs/building/better-results) is the biggest lever there is.
3. **Batch related small changes** into one message rather than sending five messages for five tweaks.
4. **Use Edit Mode for visual changes.** Deterministic style edits bypass AI entirely.
5. **Start fresh chats for new topics.** Long conversations make each request carry more context than it needs.

## Frequently asked

#### I hit the daily limit. Do I lose those tokens?

No. Nothing is deducted by the limit itself. Your remaining monthly tokens are still there tomorrow; the limit just spreads them across the period.

#### Does a failed or unhelpful response still cost tokens?

The AI work was still performed, so yes, a request that did not land the way you wanted still spends its tokens. Automatic repairs of a broken build are the exception: those are always on us.

#### What happens to my top-up balance if I change plans?

It stays. Top-ups belong to your workspace and persist through plan changes, upgrades, and ladder moves.

#### Can I see what a specific request cost?

The header tracker updates after every request, so the difference before and after a message is what it cost. A per-request breakdown is not shown today.