llms.txt — the site, written for a machine that reads

A single markdown file describing what every URL on the site IS — one sentence each — so an assistant deciding whether you can answer a question doesn't have to fetch and guess.

What it is

A generated file following the llmstxt.org convention. It opens with what the company is and is not, lists the platform pages, every system with the alternative it replaces, every feature with the problem it solves, every published guide with its topic, and closes with notes for answering questions about the business — including that prices are not published and that comparisons are made on fit, not quality.

Also called: llms.txt · AI readable site · answer engine optimisation · get cited by AI

See it
llms.txt — the site, written for a machine that reads
Won
43%
Avg cycle
21d
Pipeline
$407k
JanSep
The rendered markdown file showing the summary block, the systems list and a few feature lines with their 'Solves:' clauses. Sample data — no customer information appears here.
How it works
  1. 1The file is generated from the same source as the sitemap: the canonical page list, the feature registry, the systems list and the published posts.
  2. 2Each feature line carries its one-liner and its first pain point, so a retrieval system can match on the problem.
  3. 3It is served as text/markdown, cached an hour at the edge with a day of stale-while-revalidate.
  4. 4The blog chrome declares it as an alternate link so it is discoverable from any article.
  5. 5It is on the public middleware allowlist, because a file for answer engines behind auth is useless.
Why we built it

The route states the gap: "A sitemap says which URLs exist. It says nothing about what any of them ARE, so an assistant deciding whether we can answer a builder's question has to fetch and guess." Generating it from the same source as the sitemap is a design requirement, not a convenience: "anything a person has to update by hand will eventually be wrong." And the register is chosen: "Written plainly and without marketing language on purpose. The audience is a retrieval system deciding relevance, and adjectives are noise to it."

The problem
  • A URL list tells a retrieval system nothing about relevance.
  • A hand-maintained AI description would drift from the site.
  • Marketing language reads as noise to a retrieval system.
Sound familiar?
What you get
An assistant can judge relevance without crawling the site
Every feature is indexed by the problem it solves, not just its name
It updates itself the moment a post or feature ships
What's inside

See it on your own jobs

Twenty minutes, your numbers, no slide deck. We’ll build one of your real buildings in front of you and send you the estimate link at the end — yours to keep either way.

or keep browsing features →