Build in markdown. Publish like a site.

OpnPress is built for teams that want their website content to live in markdown, not in a visual editor.

That makes the source easy for people to edit and easy for LLMs to read, traverse, and extend. The rendered site is just the public surface. The markdown tree is the system of record.

How it works

• Write pages in content/
• Use shortcodes for structure, not page templates
• Generate static output into dist/
• Publish from the build output, with Cloudflare Pages as the default target

What the LLM sees

The site exposes source mirrors and machine-readable indexes so an LLM can follow the structure without reverse-engineering the rendered HTML.

• llms.txt
• source.md
• page.json

See the home page source mirror here: source.md

Why this matters

Search and answer engines increasingly resolve the question before a site visit happens, and more people are skipping classic search for AI tools when they want a direct answer.

• Future reported that 55% of U.S. consumers use AI instead of search engines for certain tasks. Source
• BrightLocal found that 45% of consumers now ask AI for local business recommendations, up from 6% the year before. Source
• Pew Research Center found that when Google showed an AI summary, users clicked a traditional result in 8% of visits versus 15% when no summary appeared, and only 1% clicked a cited source inside the summary. Source

That is why OpnPress ships markdown mirrors, relative links, and simple page indexes. Better source data gives LLMs better context, which gives users better answers. The content still needs to make sense when the first read happens in search, in an AI answer, or through a source mirror instead of a browser visit.

The next directory

Google Search and Maps replaced the Yellow Pages because they made it faster to find a business and decide what to do next.

AI summaries and direct LLM prompts are doing the same thing now for website and company data. Users ask an LLM for recommendations, comparisons, summaries, and next steps instead of opening ten pages and stitching the answer together themselves.

For OpnPress, that means the site is not only built to be published. It is built to be read by an LLM first, so the model can surface the right page, the right office, or the right service with less guessing and better results for the person asking.