# AI Credits and Usage Limits

:::lead
Every project has a monthly allowance of AI-assisted actions — **Ask AI** instructions and **Describe with AI** tests — on top of its execution quota. This page explains the allowance, how it's spent, and how to add more with a credit pack.
:::

AI credits cover two features, tracked separately:

- **Ask AI** — inside an open recording session, you type an instruction and the AI carries it out on the live page. See [Creating Tests with AI](/docs/building-tests/creating-tests-with-ai).
- **Describe with AI** — on the **Home** page, you describe a whole test as a plan and the AI performs it from scratch in the recorder.

Neither counts against your plan's **test execution** quota — that's a separate pool. See [Plans & Usage](/docs/team-and-account/plans-and-usage).

## Plan allowances

| | **Trial** | **Starter** | **Pro** | **Scale** |
| --- | --- | --- | --- | --- |
| Ask AI instructions | 50 in total, not renewed | 300 per 30 days | 1,500 per 30 days | 6,000 per 30 days |
| Describe with AI tests | 5 in total, not renewed | 30 per 30 days | 150 per 30 days | 600 per 30 days |

The Trial allowance is a one-time pool for the whole 7-day trial. Paid plans refill on a rolling 30-day window, independent of your billing date. Legacy and custom plans have their own limits set when the plan was created — see the AI Credits tab in your project for the exact figures.

## What counts as a unit

- **One Ask AI instruction = one unit**, no matter how many actions or steps it takes to complete.
- **One Describe with AI test = one unit** for that recording session. Any follow-up instructions you then type into that same session's Ask AI panel are charged as Ask AI instructions, not a second Describe.
- Stopping an instruction early, or the AI failing partway through, still spends the unit — the work was already done. A unit is only given back if the AI never actually started.

## Viewing and managing credits

Open **Settings → AI Credits** in your project:

- **Balance** — two cards, one per pool, showing units remaining and a status word: **Healthy**, **Running low**, **Almost out**, or **Out of credits**. A card shows **Unlimited** if your plan doesn't cap that pool.
- **Usage** — a daily usage chart over 7, 30, or 90 days, split by what paid for it (plan allowance vs. credit pack), and — for Owners — a breakdown by team member.
- **Purchase history** *(Owner only)* — every credit pack bought for the project, its status, and a link to its invoice.
- **How credits work** — a quick reminder of the spending order below.

## Credit packs (adding more)

If a project runs out mid-period, the project **Owner** can buy a one-time **credit pack** instead of waiting for the next window or changing plan. Packs come in a few sizes covering both Ask AI instructions and Describe with AI tests together; see the exact sizes and current prices under **Add credits** on the AI Credits tab.

- **Plan allowance is always spent first.** Pack credits are only used once your plan's allowance for that period is exhausted.
- **Packs expire 30 days after payment** and don't renew. Unused units are lost at expiry.
- **Buying several packs stacks them** — each keeps its own 30-day expiry, and the pack expiring soonest is spent first.
- **Packs are non-refundable** once paid, and purchases are paid through the same Razorpay checkout as your subscription (see [Billing & Subscriptions](/docs/team-and-account/billing)).
- A project must be on an **active paid plan** to buy a pack — Trial and Legacy projects need to choose a plan first.

## When you run out

Once both the plan allowance and any packs are spent:

- **Ask AI** and **Describe with AI** are disabled, with a message explaining why.
- **Owners** see an **Add credits** button right where the limit was hit.
- Other roles see a note to ask their project Owner to add credits.
- Project Owners and Admins get an in-app notification and an email the first time a pool is fully exhausted in a period.

## Using your own Claude account instead

If you connect **Claude Desktop** or the **Claude Code CLI** to TestOwl, you can run Ask AI and Describe with AI through your own Anthropic account at **zero cost to your TestOwl AI credits**. See [Connect Claude (Desktop & CLI)](/docs/integrations/claude-connect).

## Next steps

::::cardgrid
:::card{title="Creating Tests with AI" href="/docs/building-tests/creating-tests-with-ai"}
See how Ask AI and Describe with AI actually work when building a test.
:::

:::card{title="Connect Claude (Desktop & CLI)" href="/docs/integrations/claude-connect"}
Use your own Claude account instead of TestOwl AI credits.
:::

:::card{title="Billing & Subscriptions" href="/docs/team-and-account/billing"}
Manage payment details and see your purchase and invoice history.
:::
::::