Analysis · updated August 2026

Claude Design: what does design work consume?

What designing with Claude Design draws from your usage allowance, what drives that consumption up, and how it compares with Figma Make, Lovable, v0, Bolt.new and Cursor. Fully revised after the June 2026 update, which changed the pricing model fundamentally.

In shortWhat to expect from this document

Claude Design launched on 17 April 2026 as an Anthropic Labs research preview. On 17 June 2026 a major update reversed the pricing model: the separate, opaque usage allowance for design work is gone. Claude Design now counts towards the same usage allowance as your regular Claude conversations, Claude Code and Claude Cowork.

That makes the cost question both simpler and different. Simpler, because you only have one pool to watch and no second subscription limit to budget for. Different, because design work now competes with your coding and your day-to-day conversations. The question is no longer "how much Claude Design do I get on top", but "how much of my total Claude allowance goes into design". This document sets out the current facts, works through the plans, compares the known alternatives and draws a sober conclusion.

Have you read this before?

The earlier version of this article, written in late April 2026, described Claude Design as a service with its own weekly usage allowance separate from the rest of Claude. That was true then, but has not applied since 17 June 2026. If you chose a plan based on that information, it is worth reading section 01.

Looking for the subscription prices themselves?

This article is specifically about design work in Claude Design and what it draws from your usage allowance. If you want to know what Claude as a whole costs, with every plan, limit and rate side by side, read What does Claude cost? Plans and rates compared.

About the prices in this document

Anthropic bills in US dollars and does not publish a separate EUR price list. The euro amounts below are conversions at roughly €1 = $1.156 (as at 6 August 2026), excluding VAT. Note that the euro has strengthened noticeably since April 2026, so the same dollar price now costs fewer euros than in the previous version of this article. For consumers in the Netherlands, 21% VAT is added on top of the USD price. For business customers with a valid VAT number, the reverse charge generally applies at Anthropic Ireland.

01What changed about the pricing model in June 2026

At launch, Anthropic metered Claude Design separately. That produced two complaints: users could not tell how much they were getting, and consumption ran high. The update of 17 June 2026 addressed both.

Topic At launch (April 2026) Now (since June 2026)
MeteringIts own usage tracking and weekly limit, separate from the rest of Claude.One shared allowance with chat, Claude Code and Claude Cowork.
TransparencyNo public number of designs or prompts per plan.Still no number of designs, but the same familiar session and weekly limits as the rest of Claude.
Consumption per promptA point of criticism; reviewers burned through their weekly allowance quickly.Anthropic reports fewer tokens per turn and sharply fewer errors.
Paying twiceDesign on one meter, implementation via Claude Code on another.One meter for both.
Enterprise administrationOnly an on/off switch.Spend caps per organisation, group and user, plus consumption reporting and export.
StatusResearch preview.Beta, still under Anthropic Labs.

The essence: the old objection that Claude Design was "a second subscription inside your subscription" is gone. What remains is a different question, namely how you divide one shared pool between designing, coding and chatting.

02How the cost model works now

Anthropic's own wording is unambiguous:

Claude Design counts toward the same usage limits as the rest of Claude. Design activity draws from the shared pool you use for chat, Claude Code, and Cowork, so there's no separate Claude Design allowance to track.

Source: support.claude.com, Get started with Claude Design

Three consequences you do not want to miss.

First, your session limit resets every five hours. Paid plans add a weekly limit on top. There is no separate seven-day design cycle any more, but the same rhythm as the rest of Claude.

Second, design work comes at the expense of your other Claude use. An afternoon of intensive prototyping can mean you hit your limit that evening during a coding session. This is not a theoretical risk: Anthropic itself warns that Claude Code and Cowork sessions are significantly more token-intensive than ordinary chat.

Third, how much you can do within that limit depends on more than the number of designs: the length and complexity of your work, the model you choose, the effort level you set and the size of attached files or repositories all count.

What happens when the allowance runs out

You have three options: wait for the limit to reset, move to a higher plan, or enable usage credits. With usage credits you keep working at standard API rates, billed as a separate line item alongside your subscription. For Claude Opus 5 that is $5 per million input tokens (~€4.33) and $25 per million output tokens (~€21.63).

Usage credits are available on Pro, Max 5x and Max 20x. You can set a monthly spending cap, enable auto-reload and receive alerts as you approach your limit. Anyone subscribed through a mobile app has to purchase credits via the web version.

What Anthropic still does not publish

A fixed number of designs, prompts or credits per plan. That was the case in April and still is. The difference is that you no longer have two unknown pools, but one known pool with an unknown split. If you need to budget, measure your own use first. See section 05.

03Which plans give access?

Plan Claude Design access Intended for
FreeNoNot available. You need at least Pro.
ProYesExploration and occasional design work alongside normal Claude use.
Max 5xYesFive times the usage per session compared with Pro. For weekly designers or coders.
Max 20xYesTwenty times the usage per session compared with Pro. For daily intensive use.
TeamYesIncludes Claude Code and Claude Cowork. Gives access to organisation-wide design systems.
EnterpriseYes, after activationOff by default. An administrator enables it under Organisation settings, in the Anthropic Labs section.

Two changes compared with the spring. Claude Code is now simply part of Team, not only of the more expensive premium seat; the distinction between standard and premium seats is about the amount of usage, not about access to Claude Code. And on Enterprise you can restrict access to specific groups or departments through custom roles, instead of all-or-nothing for the whole organisation.

04What does each plan cost?

All prices were taken from Anthropic on 6 August 2026. The EUR amounts are indicative conversions excluding VAT.

Individual plans

Plan EUR / month
indicative, excl. VAT
EUR / year
indicative, excl. VAT
USD / month USD / year
Free€0€0$0$0
Pro (monthly)~€17.30~€208$20$240
Pro (annual, paid up front)~€14.70~€173$17 effective$200
Max 5x~€86.50~€1,038$100$1,200
Max 20x~€173~€2,076$200$2,400

Team and Enterprise

Plan EUR / seat / month
indicative, excl. VAT
USD / seat / month Conditions
Team, standard seat (annual)~€17.30$20Includes Claude Code, Cowork and Claude Design
Team, standard seat (monthly)~€21.60$25Same features, higher monthly price
Team, premium seat~€86.50$100Considerably more usage headroom per seat
EnterpriseOn requestFrom $20 per seat + consumptionSeat price plus token consumption from a shared organisation pool

Cost over a full year

Package EUR / year
indicative, excl. VAT
USD / year What you get
Pro annual~€173$200One user, all Claude surfaces, light projects
Max 5x~€1,038$1,200One intensive user who both designs and codes
Max 20x~€2,076$2,400One full-time user without daily limit worries
Team, 5 standard seats annual~€1,038$1,200Small team, shared design system, light exploration
Team, 5 premium seats annual~€5,190$6,000Small team designing and implementing daily

A team of five using Claude Design seriously therefore lands realistically between €1,000 and €5,200 per year, depending on seat type and excluding any usage credits. If you only want to explore, ~€173 (Pro annual) covers it. Because the euro has strengthened since April, these figures are lower in euros than in the previous version of this article, at unchanged dollar prices.

05What does design work cost in practice?

The honest answer: Anthropic publishes no cost per design, and any table that does is an estimate. Rather than suggest a figure we cannot substantiate, what follows is what really drives consumption and how to measure it yourself in a single week.

What drives your consumption up

  • Number of correction rounds. By far the biggest factor. A design finished in two rounds costs a fraction of one that needs eight.
  • Broad versus local changes. A chat instruction that regenerates the whole page costs more than an inline comment on one element, or a correction you make by hand on the canvas.
  • Size of the attached context. Large repositories, many screenshots and high-resolution images consume more. Link very large repositories from Claude Code instead.
  • Model and effort level. A heavier model at a higher effort level consumes more per turn.
  • Conversation length. Long sessions with a lot of accumulated context add up faster than a series of short, focused ones.

Calibrate in week one

This is the only method that reliably predicts your own situation:

  1. Work normally for a week on your current plan and note when you hit a session or weekly limit.
  2. Record what you were doing that day: only designing, or also Claude Code and long chats.
  3. If you never hit a limit, your current plan is enough. Do not enable usage credits.
  4. If you hit limits weekly, first check whether it is a matter of method: too many broad correction rounds, no design system, oversized attachments.
  5. If it still pinches with a good method, a higher plan is usually cheaper and more predictable than structural usage credits.

Which plan suits which usage pattern?

The breakdown below is our own translation of Anthropic's qualitative positioning and the 5x/20x ratios. It is a starting point for your own measurement, not a promise confirmed by Anthropic.

Plan Suits this usage pattern
ProClaude is a tool alongside your work. You design occasionally and use Claude Code little. Expect an intensive design afternoon to use up your allowance for that day.
Max 5xClaude is a fixed part of your week. You alternate designing and coding and do not want to watch the limit every session.
Max 20xClaude is your main working environment. You design daily, run long Claude Code sessions, or both.
TeamSeveral people work with a shared design system. Choose the seat type per person: standard for occasional users, premium for the designers and developers.
EnterpriseYou want spend caps, per-user consumption reporting and a controlled rollout per department.

06Comparison with other AI design tools

All prices below were taken from the vendors on 6 August 2026. Units differ considerably per platform, so the table comes first and the interpretation second. EUR amounts are indicative conversions excluding VAT. This is a fast-moving market; always check the current pricing page before making a final decision.

Tool Free tier First paid tier What you get for it
Claude Design
(via Claude Pro)
Not available ~€17.30/mo ($20)
~€14.70/mo annual ($17)
Shared allowance with chat, Claude Code and Cowork. No separate credits, no number of designs.
Figma Make
(Figma Professional)
Starter: 150 AI credits/day, up to 500/mo, without Figma Make ~€13.85/mo ($16) per Full seat, annual 3,000 AI credits/mo per Full seat. Figma Make on Full seats only.
Lovable (Pro) 5 build credits/day, up to 30/mo, plus 20 Cloud credits/mo ~€21.60/mo ($25)
~€18.15/mo annual ($21)
100 credits/mo plus the daily free credits. Unused credits roll over.
v0 by Vercel (Plus) $5 in credits/mo, max 7 messages/day ~€25.95/mo ($30) per user $30 in credits/mo plus $2 free daily credits. Consumption is token-based and therefore variable.
Bolt.new (Pro) 1M tokens/mo, max 300K/day ~€21.60/mo ($25) From 10M tokens/mo, no daily limit, unused tokens roll over.
Cursor (Pro) Hobby: limited Agent requests ~€17.30/mo ($20) Generous Agent limits and access to frontier models. Primarily a developer tool.

What this really means

Claude Design is now the only one here without its own credit system. That cuts both ways. Advantage: you are not buying a separate product and not paying twice for one finished result. Disadvantage: you cannot see what a design "cost", because it disappears into your total Claude consumption. Figma, Lovable, v0 and Bolt do give you a number, but that number is bought separately.

Figma Make is the easiest to reason about per unit. For $16 per Full seat per month you get 3,000 AI credits, and the free Starter tier has no Figma Make at all. Anyone already on a paid Figma plan needs to buy nothing extra to start.

Lovable and Bolt target complete applications rather than designs alone. If you only want a visual prototype, you are paying there for functionality Claude Design deliberately lacks: Claude Design produces no backend, but hands off to Claude Code.

v0 has become more expensive. The entry plan is now called Plus and costs $30 per user per month, against $20 in the spring. Its pricing is also token-based, so you do not know in advance what a generation will cost.

Cursor is a developer tool. For a marketer or designer who wants to "see something quickly", it is less suitable than Claude Design or Figma Make.

How predictable are the costs?

Tool Cost unit How predictable
Figma MakeAI credits per monthGood. Fixed number per seat, additional credits purchasable.
LovableCredits per monthGood. Fixed number, unused credits roll over.
Bolt.newTokens per monthReasonable. Fixed token budget, but consumption per prompt varies.
v0Credit balance in dollars, token-basedModerate. You know the amount, not what a generation costs.
CursorUsage limits per planReasonable, with automatic mode as a fallback.
Claude DesignShared Claude allowancePredictable as a fixed monthly price, not attributable per design.

Compared with April this is a clear improvement for Claude Design. It ranked last then because there were two opaque pools. Now the monthly price is fixed and known; what you lack is insight into the split within that price. For most organisations that is an acceptable trade.

07Hidden costs and points to watch

Six points that are not on the sticker price but belong on your bill or in your risk assessment.

Design work eats into your coding headroom

This is the new point to watch, replacing the old "two separate meters". Because everything comes from one pool, an intensive design session can get in the way of your Claude Code work later that week. Anthropic explicitly warns that Claude Code and Cowork sessions are significantly more token-intensive than ordinary chat. In a team, agree on priorities for busy weeks.

Usage credits can be on without you noticing

Above your limit you can keep working at API rates. On Team and Enterprise, administrators manage this through spend caps. On Pro and Max you are responsible yourself. Set a monthly cap and enable alerts before you turn it on, not afterwards.

A good design system pays for itself

The biggest saving is not in the subscription but in the method. An imported design system that lets Claude build with your own components removes the correction rounds otherwise needed to get something on brand. Count that as a one-off investment of time.

No audit logs for Claude Design

An important nuance: since the update, Enterprise does have consumption reporting, per-user spend caps and CSV export. But audit logs specific to Claude Design, showing who created, shared or exported which design, are not supported yet. So you can see what it costs, but not who did what.

No data residency

For organisations with strict GDPR or sector requirements (healthcare, government, financial services), Claude Design is currently not an option for production work with confidential data. Anthropic explicitly confirms that data residency is not supported. This point is unchanged since April.

Beta status means this can change again

The pricing model was fundamentally reversed within two months. It worked out favourably this time, but it shows how early this product still is. Factor into your risk assessment that the terms may change again.

08When is Claude Design worth the price?

Four situations in which Claude Design makes sense.

You already work with Claude Pro, Max or Team

Then Claude Design is included at no extra cost. Since June that is literally true: no separate credits, no separate limit. For exploration it costs you nothing extra. Start there.

Your design system exists in code

Claude Design imports a component library from a repository or codebase and then builds with it. That is where it leads most alternatives. Anyone who can exploit that advantage gets output that is not only better-looking but more usable.

You want to go from prototype to production without a translation step

The Claude Code integration has worked in both directions since June. That is the most concrete differentiator, and since usage became shared you no longer pay twice for it.

You do not want another vendor and another invoice

One subscription, one invoice, one limit. For organisations trying to contain their software estate, that weighs more than a few euros a month.

Three situations in which Claude Design is not the right choice.

You need to attribute costs per project or department

Because design work disappears into the shared allowance, you cannot say what a specific design cost. On Enterprise you see consumption per user, but not per deliverable. If you need to recharge at project level, a credit system serves you better.

You work with confidential or regulated data

No audit logs for Claude Design and no data residency. Wait until this is mature enough for business use, or restrict use to non-confidential material.

You have a Figma team that is already productive

Figma Make is part of every paid Full seat and fits your existing way of working. For incremental improvement you do not need to buy a second platform.

09Conclusion: a sober cost reality

The June 2026 update took the sharpest edges off the pricing policy. The main conclusion of the previous version of this article, that the entry price is misleading because a second, unknown meter runs alongside it, no longer holds. Claude Design is now simply part of your subscription, draws from the same pool as the rest of Claude, and according to Anthropic consumes less per turn than at launch.

What remains is a real but manageable objection: you cannot see what a design costs, because it disappears into your total Claude consumption. For anyone who has to recharge at project level, that is a genuine problem. For most organisations, one fixed monthly price with no credit administration is actually preferable to the alternative. Against that, the remaining limitations around administration and data storage are unchanged, and for regulated sectors they are still decisive.

Three recommendations

1. Start with your existing Pro, Max or Team subscription. Claude Design costs you nothing extra there. Measure over the first two weeks when you hit limits and why. Adjust your plan afterwards.

2. Invest in the design system first, not in a more expensive plan. Most unnecessary cost sits in correction rounds that a well-imported design system makes redundant. If it then turns out you structurally lack headroom, Max 5x (~€86.50 per month) or Max 20x (~€173 per month) is usually cheaper and more predictable than ongoing usage credits.

3. Hold off on confidential projects until Anthropic delivers data residency and audit logs for Claude Design. Until then it is an excellent exploration tool for non-confidential work.

If you want to understand not just the costs but also how it works and what the business preconditions are, read our complete guide to Claude Design. And for the broader question of what a Claude subscription costs, with every plan and limit side by side: What does Claude cost?

From cost to return

What a tool costs can be calculated. What it returns cannot, because that depends on the people using it. In practice we see the same pattern in organisations that have just started: the subscription is arranged, the invoice sits on a cost centre, and then very little happens for three months. Not because the tool disappoints, but because nobody agreed which work would actually go to it.

That is where our work begins. In the Claude training participants work with their own documents and learn where Claude stops just as much as what it can do. If your organisation runs several tools alongside each other, say Copilot for the Microsoft work and Claude for the writing, the hub covering all AI and Copilot training goes into that choice.

SourcesVerified on 6 August 2026

More control over AI tools?

Train your team to work responsibly with Claude and other AI tools

Mellaart Trainingen helps Dutch organisations use AI practically, safely and purposefully. Online or on location, always tailored to your way of working.