The Figma agent uses AI credits each time you send a prompt. The number of credits a prompt uses varies based on factors such as the complexity of the task and the amount of context the agent processes to complete it.
The practices in this article aren't only about spending fewer credits on any single prompt. The goal is to help you avoid spending credits on rework. Some upfront investments (such as writing a skill or building a generative plugin) cost credits at first, but reduce your overall spend by helping you get the result you want with fewer corrections.
For a full overview of how credits are consumed by feature, see How AI credits work. To understand how to purchase additional credits for your team, see Manage AI credits.
Key moments that affect credit consumption
Not all prompts carry the same cost. In practice, a handful of moments account for most credit inefficiency:
- An unscoped first prompt: When your first prompt is vague or incomplete, the follow-up prompts you send to correct the result cost credits. A detailed, precise starting point avoids that issue.
- Context you don't need: Attachments, connectors, and canvas selections all add to what the agent processes. Context that doesn't improve the result still costs credits.
- A long chat: The agent carries a chat's full history forward as context. As a chat grows, even simple prompts cost more.
The sections below cover these patterns and other optimizations in more detail.
Before you start working
Check your credit balance
Before you start a working session, check how many credits you have left and when they reset. Knowing your balance helps you make deliberate choices, like how much to scope into a single session or whether to invest in a reusable foundation that pays off later.
To check your AI credit balance:
- Open a file where you have edit access.
- Open the Main menu of the file.
- Select AI balance to see how many credits you've used, how many remain, and when they reset.
Note: Other Figma AI features, including Figma Make, draw from this same balance.
Learn more about tracking your individual credit usage.
Set up skills to give the agent context upfront
Skills are reusable sets of context and instructions that guide how the agent responds to your prompts. When you use a skill, you don't have to restate standing information (such as your preferences, naming conventions, or team processes) in every prompt.
This helps the agent get things right in fewer tries, which cuts down on the back-and-forth that consumes credits. You can create your own skills, or use skills from the Figma Community.
A good skill can cover:
- How to use your design system
- Naming conventions and patterns to follow
- Quality standards to aim form
- Where to find the right source of truth for common decisions
Tip: You can ask the agent to draft a skill for you, but keep in mind that generating or revising one uses credits. Once you have a skill that works, edit it directly instead of regenerating it each session.
Learn more about creating skills.
During your session
Front-load the details in your first prompt
Your first prompt acts as the brief the agent works from. The more clearly it defines the scope, requirements, and end state you want, the less you'll need to cover in follow-ups.
Every prompt costs credits, including the ones you send to fix a previous result. The most reliable way to reduce rework is to state your intent upfront.
For example, instead of prompting:
Design a settings page.
Try:
Design a mobile-first settings page with sections for Account, Notifications, and Billing, using our existing sidebar navigation pattern.
Open-ended tasks—like generating a design from scratch or building a plugin—require the agent to reason and explore, which uses more credits and time. Well-defined tasks—like applying a style across multiple components or updating text across several frames—typically cost less, because there's less ambiguity for the agent to work through.
Both are valuable ways to use the agent.
Include only the context the task needs
More context isn't always better. Every piece of context the agent processes costs credits, including context that doesn't improve the result.
When you add attachments, connectors, or frames to a prompt, bring in what the task actually needs rather than casting a wide net.
- Canvas context: Scope your selection to the frame or layer you're working on, not the whole page.
- Connectors: Point the agent to the specific document, ticket, or file that's relevant, not a broad workspace search. If you only need a short excerpt, paste it directly into your prompt instead.
- Redundancy: If the same information lives in a skill you're using, a connected file, and your prompt, the agent processes all three. Before you send a prompt, check whether any of that context overlaps and trim what's already covered elsewhere.
Ask Fig your product questions
If you have a question about how a feature works or how to do something in Figma, ask Figma's support chatbot instead of the agent.
You can chat with our support chatbot from our support hub or from within Figma by selecting the question mark icon in the bottom-right corner of any page and selecting Contact support.
Fig answers questions about Figma's products for free. The agent supports design and development work in your file. Using the right one for the job means you aren't spending credits on something that doesn't need them.
Learn more about getting support in Figma.
Over time
Start a new chat for each task
When you keep working in one chat, the agent carries that chat's full history forward as context—even the parts that are no longer relevant. The benefit is that you don't have to repeat yourself. The tradeoff is that as a chat gets longer, the agent has more prior context to process on every new prompt, which can increase credit usage even for simple requests.
Signs it may be time to start a new chat include:
- You're switching to a task unrelated to what you've been working on
- You're taking your work in a meaningfully different direction
- Simple prompts are consuming more credits than you expect
Tip: You can run multiple chats at the same time, so you don't have to wait for the agent to finish working before you start something new.
Turn repeated workflows into generative plugins
If you find yourself prompting the agent to do the same task over and over, you can turn that task into a generative plugin.
Plugins work well for well-defined, repeatable jobs, such as:
- Auditing a file for unused variables
- Flagging unnamed layers
- Applying a naming convention across components
Building the plugin costs credits, since you work with the agent to define how it should behave. However, once a plugin is built, future runs cost 0 credits.
Learn more about generative plugins.
Skills or plugins?
Skills and plugins both help you work more efficiently, in different ways:
- Skills give the agent context so it gets things right in a single prompt. This saves you the credits it would take to correct the result afterward.
- Plugins define repeatable logic upfront, so future runs don't consume credits at all.
If you run out of credits
If you use all your credits, you may be able to request more from your admin, depending on how your plan is set up. Your admin is notified and can approve or deny the request.
Learn more about requesting more credits.