For web designers, builders and small agencies
Give it the brief and two references. It researches, designs, writes the copy and builds the pages in your stack.
What's included
Researches the business, audience and competitors, decomposes references into principles, and produces the brief, sitemap, page objectives and conversion journey; marks what is known, inferred or unknown.
Turns strategy into a design direction with a stated design idea, a design system, section-level page blueprints with responsive and motion rules, and copy derived from audience, offer, proof and page intent.
Builds the site: turns a brief or a blueprint into real page files on your stack, picking vetted components first and running typecheck, lint, tests and a render before it says done.
Built from
Website strategy practice: goal, audience and offer analysis before layout, a sitemap with a stated objective and primary action per page, and a conversion journey that maps proof and objections to sections
Design direction method that names one design idea and derives type scale, grid, palette, motion and imagery from it, with an explicit list of generic patterns to avoid
Front-end engineering discipline: inspect the stack before touching it, component selection in a fixed order, and a build loop that ends with typecheck, lint, tests and a rendered check
Inside the agents
Real sections and real components, running here the way they will run on your client's site.
Motion background, glass menu, a headline that reveals itself, one action.
The tier you want chosen gets the light. A search bar with a travelling line.
A button that splits into options like liquid, and merges back.
Liquid metal on the upgrade button, the send button and the New badge.
Every issue with a page, a severity and a fix. Blockers first.
Copy and designs go to the client as cards. Approve, edit or reject.
Give it a company URL and three references. It researches the offer and audience, pulls the principles out of the references without copying them, and returns a sitemap with an objective, primary action and section plan for every page.
Every direction states its design idea in one sentence and follows it through type, grid, color, motion and imagery. No default purple gradient, no three-card sections on repeat, no copy that could belong to any company.
The Builder inspects the repository or builder first, reuses what exists and searches for a fitting component before writing a custom one. In Claude Code it runs typecheck, lint, tests and a render before it says a page is done; in Cloud it writes the code and hands you that checklist.
Add the Auditor and every page is checked at real widths for overflow, contrast, missing metadata, heavy scripts and a hero that does not say what the company does; each finding carries evidence and a fix, ordered by severity.
Write in your language and they answer in it, right to left included. What you tell them stays known next time.
One-time payment. Instant access.
FAQ
Three agents (Website Strategist, Web Designer & Copy Director, Web Builder Engineer) usable inside Agents Lab Cloud, plus the skill files, scripts and sample projects to download and run in Claude or Claude Code yourself, where the Builder can work directly in your repository.
No. The Strategist and Designer produce specifications that work for Webflow, Framer, Figma handoff or code. The Builder implements in HTML and CSS, React, Next.js, Vite or whatever the existing repository uses, and inspects it before changing anything. It does not force a framework.
No. Research is cited when web search is available; anything not researched or supplied is marked inferred, assumed or unknown and listed as a question for the client. It never invents testimonials, numbers, results or competitor facts.
No. It plans, designs, writes, builds and checks. Deploying to production, sending anything to a client and publishing content stay with you; the agents mark those items as waiting for approval.
Both. A single campaign page gets a brief, blueprint, copy and build in one session; a full site gets a sitemap, page-by-page work and, with the Auditor, a full pre-launch check at real widths.
The package is a one-time purchase and the skill files are yours. Using the agents inside Agents Lab Cloud is pay per use: $0.0325 per 1,000 input tokens and $0.1625 per 1,000 output tokens, plus $0.065 per web search. Most exchanges land between $0.50 and $3, depending on how much the agent reads and writes. Usage is charged daily to your saved card once it reaches $3, and every session is itemised in your account.