How to build client websites with Claude Code and Fable 5
Build client websites with Claude Code and Fable 5 by standardizing a repeatable pipeline: scope the brief, generate visuals in Fable 5, and let Claude Code assemble the build. The speed and margin come from reusing an indexed brand kit per client so the agent retrieves proven components instead of rebuilding each site from scratch.
Why the '$10,000 website' framing misleads you
The headline sells a number, not a method. A site is worth what the scope and the outcome justify, and no tool sets that price for you. Claude Code and Fable 5 do not make a page more valuable; they make you faster at producing it. Chase the dollar figure and you optimize the wrong variable.
The variable that actually moves your margin is repeatability. A freelancer who rebuilds every site from a blank prompt earns per hour. A freelancer who runs the same pipeline across ten clients earns per project while spending less time each round. The gain compounds because the reusable parts, your components, your intake process, your brand-kit structure, get retrieved rather than rebuilt. Focus there and the pricing conversation takes care of itself.
The repeatable client pipeline, stage by stage
Treat every client site as the same four-stage path. When the path is fixed, the only thing that changes per project is the brand kit and the copy, and both can be fed to the agent as retrievable context.
Stage 1: Scope and capture the brand kit
Start with a structured intake, not a conversation. Capture the palette, type scale, logo assets, tone, reference sites, and the concrete page list. Store this as a single brand-kit document per client. This document is not paperwork; it is the context you will index so Claude Code and Fable 5 both build against the same source of truth instead of your memory of a kickoff call.
Stage 2: Generate the visual direction in Fable 5
Use Fable 5 for what it is good at: look, feel, layout direction, and on-brand imagery. Feed it the brand kit so the output respects the client's palette and tone rather than a generic template. Lock the visual direction with the client before any build begins. Approving direction on visuals is cheaper than approving it on shipped code.
Stage 3: Build with Claude Code against real components
Hand the approved direction to Claude Code and have it assemble the site from your existing component library, not from scratch. This is where an indexed component set matters most. When the agent can retrieve your real Section, Hero, and PricingTable with their actual props, it composes the site instead of inventing markup you then have to correct.
Stage 4: Review, hand off, and template the deltas
Review composition and content, not hallucinated APIs. On handoff, note what was genuinely new about this project and fold the reusable parts back into your indexed library. Every client makes the next one faster, because the delta shrinks each time. This feedback loop is the entire point of running a pipeline instead of a series of one-off builds.
Reuse an indexed brand kit across every project with RDK
The pipeline only compounds if the reusable parts are retrievable. Index your component library, layout patterns, and each client's approved brand kit as encrypted private chunks on the RDK network. Claude Code searches those chunks before it queries the model, so it pulls the real component signatures and the exact brand tokens instead of guessing.
This is stacked retrieval applied to agency work. Your private vault answers most of the "what does this client's system say" and "how do we build a pricing section" questions. The public network can add proven patterns you have not built yourself. The model only handles the genuinely novel composition for this specific site. Because those repeated answers are retrieved rather than regenerated, token spend on them drops 80 to 90 percent, and each site lands on-brand on the first pass.
The practical payoff is time. A blank-prompt build spends your review cycles fixing invented components and off-brand colors. A retrieval-grounded build spends them on layout and copy, the parts a client actually pays for. The brand kit that lives in a shared drive cannot help your agent. The brand kit indexed as chunks becomes the source the agent reads on every request, across every project you take on.
Turn your reusable components into USDC income
The parts of your pipeline that are not client-specific are an asset. A clean pricing-section pattern, a tested contact-form component, a well-structured landing-page starter, none of these expose a client's brand or content. Publish them as public chunks on the RDK network. When another builder's agent retrieves one, you earn USDC per retrieval.
This is a second revenue line that does not compete with your billable work. You build the component once for a real project, strip the client specifics, and publish the generic pattern. The same chunk that speeds up your next build now earns every time someone else's agent pulls it. Publish once, earn on every retrieval, without maintaining a marketplace or chasing a commission. Payments settle on Base in USDC, so the earnings accrue as the network uses your work.
The framing matters more than any headline price. You are not selling one expensive website. You are building a retrievable body of work that speeds up your own delivery and pays you when the wider network reuses it. That is a more durable business than any single project fee.
Frequently asked questions
- Can Claude Code and Fable 5 really build client-ready websites?
- Yes, when you supply context instead of a blank prompt. Fable 5 sets the visual direction and Claude Code assembles the build from your real components. The output is client-ready when the agent retrieves an indexed brand kit and component library, so it composes on-brand pages rather than inventing markup and colors you then have to correct.
- How do I make the workflow repeatable across clients?
- Fix the pipeline and vary only the brand kit. Capture each client's palette, type, and assets as a structured document, index it alongside your reusable components as RDK chunks, and run the same four stages every time: scope, generate direction, build, hand off. The reusable parts get retrieved per project, so each new client is faster than the last.
- Should I promise clients a specific price like $10,000 per site?
- No. A price should follow scope and outcome, not a tool. Claude Code and Fable 5 make you faster; they do not set value. Price on the work the client needs and let your indexed pipeline compress delivery time. Chasing a headline dollar figure optimizes the wrong variable and sets expectations you cannot control.
- How can reusable components earn me money?
- Publish your generic, non-client component patterns and starter kits as public RDK chunks. When another builder's agent retrieves one, you earn USDC per retrieval, settled on Base. You build the component once for a real project, strip the client specifics, and publish the generic version. It speeds up your next build and pays you when the network reuses it.