Case study

A tool site designed, built and shipped by an AI agent pipeline

nocodecsv.com — a CSV and data-tool site for people who work with spreadsheets. Every page, article and technical layer was produced by an agent pipeline I operate: specification, generation, deterministic QA, then deployment. No hand-coding, no template purchase.

What shipped — verified, not claimed

85pages in the sitemap
71published articles
14tool pages
85 mshome page response (measured)

Figures measured on the live site. No traffic or revenue numbers are claimed here — this page documents the build and the technical layer, not commercial results.

The part most sites skip

Agent-ready layer

A normal website is readable by humans. These files make it usable by AI agents — search assistants, procurement agents, research tools.

01

llms.txt

A machine-readable summary of the site for large language models — what it offers, where the key pages are.

02

agent-tools.json

A declared tool manifest: the site's capabilities in the shape an agent can call, not the shape a browser renders.

03

WebMCP

Browser-level tool exposure, so an agent visiting the page can act on it instead of only reading it.

04

Structured data & canonical layer

JSON-LD, canonical tags, sitemap and Open Graph — the vocabulary search engines and assistants rely on.

How it was built

Specification first, then generation, then deterministic checks

01

Interrogate the requirement

Before any build: force the vague idea into a written specification — who it is for, what it must do, what it must not claim.

02

Generate the site

Pages, copy and structure produced by the pipeline against that specification.

03

Deterministic QA

Automated checks on every release: link integrity, metadata, structured data, indexing signals. A page that fails does not ship.

04

Ship, and keep shipping

Deployment, sitemap updates and indexing pushes run on a schedule. Monitoring stays silent unless something breaks.

What this means for you

The same pipeline, applied to your site

If you need a site that AI assistants can actually use — not just read — this is the work: an audit of what your site currently exposes, the agent layer added on top, and a verification report you can keep.

Ask about an agent-readiness audit

Get a Quote