# Spend fewer credits

> Get the same result for fewer credits - the right build quality, one part at a time, a new chat per task, lower Effort and Context, and a cost before big work.
> Source: https://playcode.io/docs/build/spend-fewer-credits - last reviewed 2026-10-09.
> Ask Playcode's agent: "Before you change anything, tell me what this involves and roughly what it will cost." It names the work and a starting cost in credits, a floor rather than a total; it cannot see your balance or what earlier work cost.

Every step the agent takes uses credits, and three things set the cost: the build quality, how much of the project a request touches, and how long the chat is. Choose the lowest level that does the job, ask for one part at a time, and start a new chat for each new task.

## Do it yourself

1. Choose the build quality for the task: **Economy** for simple edits, fixes and basic pages, **Standard** or **Ultra** for hard work. Ultra "may use credits several times faster than Economy or Standard". See [Models and modes](/docs/build/models-and-modes).
2. Ask for one part at a time and name it: "Add a fade-in to the pricing cards", not "Add animations to the site". A small request reads fewer files, and it is cheap to check and to undo.
3. Before a big job, ask what it will cost. The agent names the work and a starting cost in credits; the final cost can be higher.
4. Start a new chat for each new task: in the left panel, under the project, choose **New Chat**. Each step of the agent carries the chat so far, so every message in a long chat costs more. To see where a chat stands, open the project menu (☰ and the project name): the bottom shows **Spent in this chat** and **Context**. At 80% the app says: "Long chats reply slower and cost more - start a new chat for the next task."
5. Keep **Context** at **High** or lower, and choose a lower **Effort** for simple changes. The app's advice: "use the lowest setting only for very simple changes, and the highest for the toughest tasks." Both dials are in **Choose build quality**.
6. Watch the first steps of a run, and stop it as soon as it goes the wrong way: choose the stop button (**Stop generation**), then say what you want instead.
7. When the same fix fails two or three times, stop asking for it again. Go back to the last good point with [Undo and checkpoints](/docs/build/undo-and-checkpoints), then describe the problem differently or attach a screenshot.
8. Check what the work cost in **Settings**, **Usage**. Workspace owners and admins see each request there; other members see who pays.

## If it doesn't work

### Credits still go fast

Open **Settings**, **Usage** and look for what used them. Besides the agent's work, credits pay for generated images, video and audio, background tasks, and the hours a Cloud app's computers are awake: see [Run a Cloud app](/docs/publish/cloud-apps). See also [Credits went faster than expected](/docs/help/credits-went-faster).

### "You're out of credits. Add more to keep building."

The workspace has no credits left, and the agent stops. On Pro or Max, add credits: see [Top-ups and auto-recharge](/docs/account-and-billing/top-ups-and-auto-recharge). On the Free plan, the app offers an upgrade.

### "I got stuck in a loop and stopped to save your credits."

The agent repeated itself and stopped. Send the request again in other words. If the chat is on Economy, try Standard: the app warns that on complex tasks Economy "can sometimes repeat or get stuck in a loop".

## Limits

- The agent cannot see your balance or what earlier work cost. Its estimate is a floor, not a total.
- A stopped run keeps the credits it already used. **Undo & resend** runs your message again and uses credits again.
- On the Free plan, only Economy is available.
- Where your credits went, by day and by request: see [Where your credits went](/docs/account-and-billing/where-your-credits-went).