Zorto is a static site generator built for workflows where humans and AI agents collaborate on websites.

Design principles #

  • Explicit contracts. One config.toml and markdown files with TOML frontmatter. No implicit file conventions or magic directories.
  • Markdown-first content. AI models understand markdown natively. No proprietary content formats.
  • Predictable output. Same input produces the same output. No runtime state, no database, no visitor-side side effects.
  • Strings in config, not templates. Text content belongs in config.toml or frontmatter, not hard-coded in HTML.
  • Executable code blocks. Dynamic output at build time, baked into static HTML.

Build-time validation #

Zorto validates your site during the build:

WriteContent and config
→
BuildCompile the site
→
CheckValidate structure and links
→
OutputHTML or clear errors

Current checks: internal link validation (broken @/ references), template syntax, missing files, invalid configuration. Planned: mobile-friendliness, accessibility, semantic HTML validation.

llms.txt #

Zorto generates an llms.txt file at build time by default — a machine-readable index of your entire site. An agent can read a single URL and understand the site’s structure and content without crawling.

This creates two files:

  • /llms.txt — links to markdown versions of every page
  • /llms-full.txt — the full content of every page in a single file

This is for consumption, not editing. Agents that need to modify a site work directly on the filesystem.

Markdown file generation #

When enabled, Zorto generates a .md version of every page alongside the HTML, accessible at the same URL with a .md extension:

generate_md_files = true

Further reading #