llms.txt and Markdown pages

llms.txt is a convention for publishing documentation in a form that language models can read. The NotForm docs publish three views.

The index

llms.txt is a short Markdown file that lists the documentation pages with a short description of each. It is the cheapest way for an assistant to learn what exists, and it is small enough to load at the start of any task.

Typical use: give the URL to an assistant and ask it to pick the pages it needs.

Read https://notformdocs.vercel.app/llms.txt, then read the pages about array fields and validation, and build a repeatable contacts form.

Read a single page

Add .md to a documentation page's path to get that page as Markdown. Use the section and page names from the URL in your browser.

https://notformdocs.vercel.app/docs/components/field.md
https://notformdocs.vercel.app/docs/guides/validation.md
https://notformdocs.vercel.app/docs/recipes/server-errors.md

Loading focused pages keeps the assistant's context small and its answers specific. This is the preferred way to use the docs with an AI tool.

The full documentation

llms-full.txt contains the whole documentation in a single file. Use it when:

• the task spans many sections, such as a migration;
• the tool cannot follow links or fetch several pages; or
• you want to load everything once and ask many questions.

It is large. For a single question, the index plus one or two pages is faster and cheaper.

Which should I use?

Using the files with common tools

• Chat assistants: paste the URL into the conversation and ask the assistant to read it.
• Editors and coding agents: add the URL as a documentation source in the tool's settings, or reference it in your project's agent instructions.
• Your own scripts: fetch the .md URLs directly. They are plain Markdown.

Keep in mind

• The files describe the latest published documentation. If your project pins an older NotForm version, check that the API you rely on exists in your installed package.
• The index lists pages, not behavior. Always read the page before writing code against it.