Figma Make Explained: Prompt-to-App, Seats and Credit Costs

Figma Make is Figma's prompt-to-app tool: you describe what you want, attach a design if you have one, and Make builds a working, code-backed prototype or web app you can click through, edit and publish. It is the fastest way to see a design behave like software. It is also the main reason Figma AI credits run out. Both facts matter.

This page covers what Make does, who can use it, what it costs in credits, and where it fits, from Figma's own pages read on 9 October 2026.

What Figma Make is

Figma calls Make an AI-driven, prompt-to-app tool that turns ideas and existing Figma designs into functional prototypes, web apps and interactive UI ^[104]. It has two areas: a chat where you prompt the model, and a preview where you interact with the result and edit its code ^[105].

It is for previewing designs in functional form, turning proposals into prototypes, and exploring several solutions to a problem ^[105]. Figma describes it as a code editor for prototyping on your codebase or from scratch ^[103].

## A short history

Figma introduced Make on 7 May 2025, running at launch on Claude 3.7 Sonnet ^[65]. In September 2025 it added copying a Make preview to the design canvas as editable layers ^[68]. In December 2025 Figma said its AI features, Make included, were moving out of beta ^[6].

In 2026 Make gained OpenAI's GPT-5.6 on 9 July ^[86] and, from 28 May, a beta that works in a team's production codebase through the desktop app ^[67].

What Figma's own study found

In August 2026 Figma's data science team published a randomized controlled trial with 100 participants working through tasks with and without Make ^[69]. It reported design work 20% faster and 16% easier with Make, and the largest gains for product managers: 23% faster and 37% easier ^[69].

This is the vendor measuring its own product. Read it as such. It is still more than most AI tools publish, and the method is described in the post.

## Which model ran Make, and when

At launch Make used Claude 3.7 Sonnet ^[65]. By June 2025 Figma's tips post said Make used Anthropic's Claude Sonnet 4 ^[70]. In 2026 users choose: Auto by default, Claude Opus, Gemini Flash, and since July OpenAI's GPT-5.6 ^{[106] [86]}.

The model affects cost as much as quality, which is why Figma's credit guide treats model choice as a task-level decision ^[106].

Who can use it

Full functionality is for Full seats on Professional, Organization and Enterprise; Dev and Collab seats and Starter users can explore some features ^[103]. K-12 Education users do not get Make; higher-education and bootcamp users do ^[104]. Figma for Government includes Make while other AI features are unavailable ^[103].

On Organization and Enterprise, Make works only if AI features are enabled ^[104].

## What each seat can do

A Full seat can create Make files in folders and teams, share them and publish them ^[104]. Dev, Collab and View seats can create Make files only in drafts, and cannot share or publish them ^[104].

Exporting code is open to Full and Dev seats; attaching designs, point-and-edit, style context and the code editor work for other seats only in drafts ^[104]. In practice, Make is a Full-seat product with a trial for everyone else. Budget seats accordingly.

How a session works

You attach context by clicking Add context and pasting a design URL, or by pasting frames and components straight into the chat; images and text files can be dragged in ^[105]. Designs must be frames or components to work as attachments ^[104].

Do not paste a file link into the prompt itself. Figma warns the model may try to interpret the URL rather than the file, which usually produces an unexpected result ^[104]. Attach, do not link. Figma also suggests attaching one design per prompt when several feed the same app ^[104].

## Editing without prompting

You can also paste your own code into Make's code editor and keep iterating through conversation ^[104]. Make offers an edit tool that points at an element in the preview and opens a Figma-style properties panel, a draw tool for marking up the preview, annotations for element-level instructions, and version history ^[105]. You can also run several chats at once in one file ^[105].

These matter for cost. Figma's own advice is to use point-and-edit and direct code edits for small changes, because they bypass the model ^[106].

Choosing a model

Make offers several models. Figma's guidance: Auto, the default, balances quality and cost; Claude Opus suits complex or exact work such as a detailed replica of a design; Gemini Flash suits small styling changes and is the most credit-efficient ^[106]. GPT-5.6 was added in July 2026 ^[86].

Each model has its own pricing, and some consume significantly more credits than others ^[94].

What it costs in credits

Make draws on the same AI credits as every other Figma AI feature. Figma's examples, using its default model as of February 2026: changing a font costs about 30 credits or more, making a design interactive about 75, and generating an app from scratch about 100 or more ^[94].

A Professional Full seat includes 3,000 credits a month; Dev, Collab and View seats and Starter include 500 ^[94]. By our arithmetic, 3,000 credits cover about 30 from-scratch apps at Figma's estimate.

Why the cost is hard to predict

Credit use depends on the model, the complexity of the task, and how much context the model reads, including attached files, libraries, connectors and chat history ^[94]. Because the model decides what to do, Figma says it cannot predict a prompt's cost in advance; you see it afterwards by hovering over the credit icon ^[94].

Undoing an action does not refund its credits ^[94]. Auto fix and Fix for me do not use extra credits; a new prompt asking for the same fix does ^[94].

Plan mode

Plan mode writes a structured plan before any code. The planning step costs credits, and the build after approval costs more ^[94]. Figma argues the upfront cost is often far less than a build that has to be corrected or started over ^[106].

Use it for complex tasks you cannot scope in one prompt. Skip it for small edits. Plans cost credits.

Prompting advice from Figma

Figma's designer advocates recommend front-loading the first prompt with the task, the context, the key design elements, the expected behaviors and the constraints such as device or layout ^[70]. If a request fails, rephrase it concretely: instead of asking to vertically align two elements, ask to add 16px of space between them ^[70].

They also suggest starting a fresh Make file after too many adjustments, carrying over what the first attempt taught you ^[70].

How to spend fewer credits

Figma's credit guide names three expensive moments: a vague first prompt, the wrong model for the job, and a long prompt history that the model rereads every time ^[106]. Start a fresh file when old context no longer helps, and duplicate files to branch, which costs nothing ^[106].

Prefer a screenshot over a full frame when pixel accuracy does not matter, and choose either an npm package or a Figma library as context rather than both ^[106].

Guidelines and Make kits

A guidelines.md file is read automatically on every prompt, so standing instructions do not need repeating ^[106]. Figma suggests a short parent file pointing to task-specific files, and editing it by hand rather than regenerating it ^[106].

Make kits go further: each bundles an npm package, a Figma library and a guidelines file, giving Make your design system from the first prompt ^[106]. Kits cost credits to set up and save them later ^[106]. That trade usually pays.

Make kits in practice

Make kits are available to Full seats on paid plans, and organization admins can approve published kits and enable them by default ^[107]. A kit can include npm packages for code context, variables and styles from published Figma libraries, and guidelines on how Make should use them ^[107].

Large packages or libraries can take a few minutes to import ^[107].

Templates

Teams can publish a polished Make file as a template, set guidelines for what others may change, and let teammates start from it ^[105]. Figma's credit guide recommends templates for hero screens and common flows, so nobody regenerates the same base ^[106].

Backends and live data

Make can add a backend to store data across browsers, provide login screens, save content and keep secrets and API keys ^[105]. It can also search the web or fetch a specific URL to ground an app in real data ^[105].

Connectors give Make access to tools such as Notion, Linear or Atlassian; Figma advises pointing them at a specific document rather than a whole workspace, because every retrieved item costs credits ^[106].

Community designs and attribution

Make can attach designs from the Figma Community to a prompt. When it does, it lists them above the prompt box and creates an Attributions.md file that ships with the app ^[232]. Paid Community resources can be used by pasting their frames into the chat ^[232].

Creators who do not want their Community files found by Make can opt their whole account out of search, but not individual files ^[232].

Publishing

Publishing puts the app on the public web at a dedicated URL that anyone with the link can open; it is not indexed by search engines by default ^[104]. Published apps are hosted on Amazon Web Services, with domain and routing served by Cloudflare ^[104].

Only Full seats on paid plans and Starter users can publish, and Starter users only if they also publish to the Community ^[104]. You are responsible for what you publish ^[104].

Custom domains and controls

Make and Sites share a custom-domain allowance: 10 on Professional, unlimited on Organization and Enterprise ^[101]. Organization admins can switch website and app publishing off for the organization ^[104].

Viewers of a published app see the app only, not the conversation or the Make file behind it ^[104].

Who owns what Make produces

Figma's AI terms treat prompts and generated output, code included, as Customer Content: the customer keeps all right, title and interest in both ^[56]. The same terms say output is not verified by Figma for accuracy, and that Figma gives no warranty for it ^[56].

Content training applies too. When it is on, Figma may use Customer Content to train its models ^[56], and it is on by default for Starter and Professional teams ^[99].

Limits and errors

Beyond credits, Make has a per-user prompt limit: when it is reached, Make shows the date it resets ^[233]. Large attachments and oversized frames fail, and Figma's fix is to split designs into smaller frames ^[233].

Those errors cost time rather than credits, but they shape how you prepare designs before attaching them. Smaller frames work better.

Collaboration

Make files created on Starter or by Full seats are multiplayer, and people with view access can comment ^[104]. Files created by Dev, Collab or View seats cannot be shared without upgrading to a Full seat, so nobody else can see them ^[104].

If a product manager drafts in Make on a Collab seat, the draft stays private until someone with a Full seat rebuilds or the seat is upgraded.

Embedding Make prototypes

A Make prototype can be embedded in Figma Design, FigJam or Figma Slides by pasting its Make file link; it plays inside the file, and stopping it resets it ^[234]. Embeds work on all plans, but viewers need access to the source Make file ^[234].

Others in the file do not see you interacting with the prototype, and playing many embeds at once can slow a large file ^[234].

From Make back to design

On paid plans you can copy a Make preview as design layers and paste them into Figma Design ^[104]. Changes in Design do not flow back to Make, and the layers are not tied to your design system ^[104].

Figma also bought the technology behind the html.to.design plugin, which turns live prototypes or HTML into editable frames, to build future Make features ^[68].

Make on your own codebase

Since May 2026 a beta in Figma's desktop app lets Make work in a team's production codebase ^[67]. Figma says designers who already have access to their company's codebase are best suited to it for now ^[67].

During the beta, prompts and actions on a local codebase consume no AI credits; standard credit use starts at general availability ^[3].

Third-party content

Make may pull fonts, code packages or images from the internet into an app, and Figma says you are responsible for having the rights to all content before publishing ^[105]. Published apps can request microphone and camera access, though the preview cannot ^[104].

What users report

In our 100-review Trustpilot sample, 15 reviews concern AI credits or Make, almost all at one or two stars ^[167]. One reviewer said support told them "credits are consumed even when generations fail, time out, or produce no visible output" ^[167]. Others describe a monthly allowance used up in days on a single project ^[167].

A minority are enthusiastic: one three-star reviewer called Make fantastic and the AI surprisingly powerful ^[167].

Make against Lovable and v0

Lovable and v0 are the closest alternatives. Lovable publishes example prompt costs, such as 0.50 credits to make a button gray, and shows each message's cost afterwards ^[176]. v0 has a free plan of seven messages a day and paid plans at US$30 and US$100 per user a month ^[181].

Make's advantage is design context: frames, libraries and Make kits from the same Figma account ^[19].

What Make is not

Make is not Figma Design with AI attached. Figma's FAQ says Design tools are not directly integrated into Make; you use the edit tool, prompts and the properties panel instead ^[104]. Voice prompting needs microphone access ^[104].

It is also not unmetered. Seat credit limits have been enforced since 18 March 2026 ^[6], and a per-user prompt limit can pause work until a stated reset date ^[233].

Who should use Figma Make

Designers and product managers who already work in Figma and want clickable, realistic prototypes from their own designs. Teams with a design system in npm who can invest in a Make kit. Prototype builders who can budget credits.

Who should not

Anyone expecting predictable costs per task, teams without a Full seat for each Make author, and K-12 classrooms. Each hits a wall.

## Verdict

Figma Make is the best prompt-to-prototype tool for teams already in Figma, because it starts from their designs and system. Treat it as a metered service: set per-user credit limits, write a guidelines file, choose the cheapest model that works, and measure a few typical tasks before rolling it out.