The first hour, guided
An onboarding skill inside the plugin leads the first hour with questions, not a form. Zero live hours from me per customer, and the guidance is written in my voice inside the skill.
Nobody builds your portfolio here.
You walk out with the keys.
Promtfolio walks designers into Claude Code for one guided hour, and they walk out with a live portfolio page they built themselves. The code sits in a folder on their machine. One payment, no subscription, and no reason to come back to me.
Designers walk out with a live portfolio page they built themselves, and the code stays on their machine.
2026.
If you already work with AI, your portfolio is already written. It just is not on a site.
Designers working with Claude, Cursor or Lovable produce documentation without noticing. The prompts, the decisions, the dead end they backed out of, the moment they changed direction. Everything a good case study is made of already exists, scattered through chat history.
What is missing is not content, it is extraction. Meanwhile portfolio platforms sell the exact opposite: a template, a monthly subscription, and a site that lives on their servers.
The person I am talking to is not a designer looking for an easier way to build a website. It is someone who already picked AI as their primary tool, and has not yet found a way to show it.
Promtfolio does not build you a site. It walks you into Claude and walks you out able to build for yourself. The portfolio is your first project, not the product.
The value proposition, written 10 August 2026
I built a whole SaaS, then froze it
The first version was a fully hosted product. Analyze, preview, publish to promtfolio.com/username, monthly subscription in two tiers. The code was written. The infrastructure ran on Vercel and Supabase. In August 2026 I froze all of it.
The reason was not technical. The product contradicted its own promise. If your site is hosted by me, you did not buy it, you are renting it from me. A subscription is something you can lose, and the entire pitch rested on ownership.
What stayed alive from the old version is the demo, because it shows what you get before I ask anyone for an hour of their time. It was an expensive decision. The code is still sitting there and I am not developing it, and I still think it was the right call.
The model I froze
Monthly subscription, two tiers
The site hosted by me, on my domain
Stop paying and the site goes down
Every change routed through the platform
I am the bottleneck, forever
The model that shipped
One payment. Nothing for us to renew.
Static HTML and CSS in a folder on your machine
Your domain and your hosting, on Cloudflare's free tier
Cancelling your Claude subscription breaks nothing
You change it yourself, whenever you feel like it
Four things that keep working after the hour ends
The plugin, seven skills
Start, Extract, Build, Deploy, Audit and more. Skill files are memory you can read: they stay in the project, you can edit them, and run them again in six months without remembering a thing.
BRAND.md, your own language
A brand file created during onboarding, covering typography, color, illustration, animation, grid, spacing, accessibility and microcopy. The agent flags requests that drift from it, and you approve or drop it. A guardrail, not a gate.
Learning center and a free community
Five videos at the friction points, a written guide per skill, and an FAQ. A support agent answers first, and only refunds, complaints and a failure that came back after two tries reach me. Every future skill update is included.
Five stages, from install to a page that is live
Start
Install and calibrate. The agent learns you and your work
Extract
Pull the case study out of the workflow you already have
Build
Build the pages against BRAND.md, with live feedback
Deploy
Go live on your domain, in your account
Audit
Mobile, accessibility and speed checked before launch
Ownership is not a line on a sales page,
it is architecture
Anyone can print "your code" on a banner. It is much harder to build a product where that stays true after the customer stops paying everyone, us included. Three structural decisions make it real.
The output is static HTML and CSS. Cancelling a Claude subscription does not take the site down, because the site never needed Claude to exist, only to get built.
Every site ships with manual content management. A content folder and a local editor, the old fashioned route, so you can add a project or swap an image by hand five years from now.
And anyone who wants to keep going without Claude gets a bespoke answer from us: the hosted studio, the same code I froze, opens up for them. The test that kept me honest across the whole product: what happens to my customer on the day I disappear.
A product that is mostly text files you can open and read
There is no black box here. Every stage of the onboarding is a Markdown file sitting in the customer's project. She can open it, see exactly what it asks the agent to do, change a word, and run it again.
--- name: extract description: Pull a case study out of the work you already did with AI. --- # Extract Ask for the project folder, then read what is already there: prompts, commits, the dead ends, the moment the direction changed. ## Rules - The designer owns the taste. Never write the opinion for her. - One decision per section. If it has no "why", it is not a decision. - Quality bar: meravish.com. If the page is thinner than that, the questions were not good enough.
Claude Code
The platform the onboarding runs on. Depth in one before breadth across many.
Skills in Markdown
Every stage is a readable file. Open it, change it, run it again.
Cloudflare Pages
Hosting on the free tier, in the customer's account and not in mine.
Static HTML and CSS
No build chain and no framework that breaks in two years.
Content folder and a small builder
Manual content management that keeps working with no AI at all.
Support agent
Fed by the same skills and guides, answering first in the community.
A one woman product has to run without the one woman
I do not have the capacity to run a community, and I do not have the capacity to personally walk every customer through this. That is not a disclaimer, it is the constraint that shaped the product. Everything is built around a fixed order: self serve first, agent second, me last.
The guidance lives inside the skills instead of inside a meeting, and the community is run by an agent fed by the same guides. Three things reach me and nothing else: a refund, a complaint, and a failure that came back after the agent tried twice.
Alongside that, a learning loop is planned: friction points collected with consent and without personal content, and a weekly agent proposing changes to the skills. The idea is that the product improves from everyone who passes through it, without me reading every conversation.
Live, at launch pricing, with a real date
The Founders page is up. Anyone who joins by 31 October 2026 is a Founders Member at the launch price of $49, and the status is permanent: every future skill update is included at no extra cost. After that date the price goes up.
The low price is a decision, not a compromise. A product for many people with a large community is worth more to me than a high margin on a handful of customers, because the community is what brings the next customers.
What is still unproven: how many of them keep building on their own after that first hour. That is the only number that says whether the product did what it promised, and it is exactly what the learning loop is meant to see.
Promtfolio
The product is live and you can walk into it. The Founders page is open until 31 October 2026, one payment, and the keys stay with you.
See Promtfolio (opens in a new tab)promtfolio.com