---
url: https://pencel.ai/docs/guides/managing-costs.md
description: >-
  Understand Pencel credits, see what you spend, and set spending limits for the
  whole app, for each room and for each agent.
---

# Control what you spend

Pencel pays for AI work with **credits**. This page explains what uses them, where to see your spending, and how to set limits so nothing surprises you.

## What credits are

One credit is $0.10 of AI usage, so 10 credits = $1. Credits pay for the work Pencel does with an AI model: reading, sorting, scoring, drafting and answering your messages.

* **What uses credits:** messages you send in rooms and on Home, scheduled work that finds something to do, drafting documents, and building or testing workflows.
* **What doesn't:** a room's background check that finds nothing new uses no credits.

### How many you get

| Plan | Credits |
|---|---|
| **Free** | Free starter credits worth $5 when you verify your email and allow anonymous usage data. They expire after 30 days. After that, use your own API key. |
| **Solo** ($20/month or $200/year) | 80 credits a month, about ten agent runs |
| **Pro** ($99/month or $990/year) | 450 credits a month, about a hundred agent runs |

Plan credits reset every month. On a paid plan you can buy more at any time: **Top Up** adds 40 credits for $19. See [Accounts, plans and credits](/getting-started/accounts).

If you use **your own API key**, requests go straight to that AI provider, and the provider bills you for them directly.

## See what you're spending

Go to **Settings ▸ Cost & Usage**. You'll see:

* **Monthly Budget**: your app-wide limit, if you set one (see below).
* Totals for **Today**, **This Week**, **This Month** and **All Time**, in credits.
* **Daily Spend**: a chart of the last 30 days.
* **By Task**: which kinds of work used the most credits in the last 30 days.
* **Recent Workflow Runs** and **Recent Chats**, with what each cost.

For your balance, go to **Settings ▸ Plan & Billing**. On a paid plan, **Credit Balance** shows the credits you have left and when any of them expire. **Top Up Credits** is underneath.

To see what a single run cost, open **Studio ▸ Jobs**. The **All runs** table has a **Cost** column, and each run log ends with its total in credits.

## Set an app-wide monthly limit

1. Go to **Settings ▸ Cost & Usage**.
2. Under **Monthly Budget**, click **Set limit**.
3. Enter an amount in dollars (for example, $20 covers 200 credits of usage) and click **Save**.

The card then shows how much you've spent and a badge: **Within budget**, **80% used** or **Limit reached**. When you reach the limit, Pencel stops doing AI work until the next month starts or you raise the limit. Use **Edit limit** to change it, or **Remove spending limit** to go back to no limit.

## Give a room its own budget

A busy room can have its own monthly limit, so one room can't use up everything.

1. Open the room and click its name to open its settings.
2. Go to **Advanced ▸ Monthly budget**.
3. Enter a number of credits. Leave it blank for **No limit** (the app-wide limit still applies).

Under the field, the room shows what it has spent this month, how much is **reserved** (set aside for work in progress) and how much is **left**. When less than 20% is left, you'll see a warning.

When a room runs out, its scheduled work pauses. It doesn't fail. A banner above the room's message box explains why, with an **Open Advanced** button. The room picks up again on its own when the month resets or when you raise its limit.

## Cap what one agent can spend on a single run

1. Go to **Studio ▸ Agents** and open the agent.
2. Open the **Knowledge & Guardrails** tab.
3. Under **Guardrails**, enter a dollar amount in **Max Cost per Run**. Leave it empty to use the standard limit.
4. Click **Save**.

If a run reaches the cap, the agent stops. The run is marked as finished with problems (**Warnings** in Jobs) so a half-finished result never looks complete.

## Tips to spend less

* **Pick sensible check intervals.** An inbox room checking every 30 minutes is plenty for most people. Something that changes slowly can check daily. See [Run work on a schedule](/guides/scheduling-automation).
* **Turn off routines you don't read.** Switch them off in the room's **Schedule ▸ Routines**, or pause them in Studio.
* **Use a lighter model for simple work.** Sorting, tagging and short summaries don't need the most capable model. A lighter model such as Gemini 3.8 Flash costs much less. Set it as an agent's **Preferred Model**, pick it in the message box, or change **Settings ▸ General ▸ Default Model**. See [Supported models](/reference/supported-models).
* **Keep "always on" knowledge short.** Knowledge documents set to **Always** go into every piece of work. Set reference material to **Smart** so it's only used when it's relevant. See [Knowledge](/concepts/knowledge).
* **Check new routines early.** Look at the cost of the first few runs of a new workflow in **Studio ▸ Jobs**. Real numbers beat guesses.
* **Use your own API key** if you'd rather pay the AI provider directly.

## Related

* [Accounts, plans and credits](/getting-started/accounts)
* [Supported models](/reference/supported-models)
* [Run work on a schedule](/guides/scheduling-automation)
* [Check on scheduled work](/guides/monitoring-runs)