Check your llms.txt the way agents and parsers read it

Paste an llms.txt. Your browser checks it against the llmstxt.org format and against the proposal's own reference parser, and compares its links with your sitemap or file tree - free, and nothing is uploaded. A paid run writes a new file from your page list or brings an existing one up to date, and the page checks every link it returns.

The examples replay saved model runs - the whole page, free.

1. The llms.txt (the file the update edits)

MarkdownIts address

Drop an llms.txt here, or

2. The page list (optional - compared with the file's links)

sitemap.xml, one URL per line, Markdown links, tree output or repository paths

Drop a sitemap.xml or a list here, or

No page list yet?

The page cannot fetch your site. Any of these gives a list you can paste or drop here:

A site: open https://your-site/sitemap.xml and save the page.
A repository: git ls-files '*.md' '*.mdx'
A docs folder: tree docs or find docs -name '*.md'

Paste an llms.txt, or load an example.

Check (free, in your browser, no AI)

title / summary

sections / links

findings

reference parser

size

Findings (numbered for the update)

What an agent reads (the file's structure; flagged links are marked)

Plain outline

What the reference parser returns

The proposal's parser (llms_txt miniparse) as this page reproduces it: the title, the summary, the details and each section's links - or the line where it gives up.

Starter draft (made by the page from the page list, no AI)

Pages grouped by their path, skippable ones left out, titles taken from the URL. The notes are left for you - or for a create run.

Free: no account, no run.

3. What changed (optional; give full URLs for new pages)

Check a file first to price the run.

Raw model reply

The reply did not parse as the structured result, so it is shown as it arrived.

Your recent runs

What this does, and what it does not

An llms.txt is a short Markdown file at /llms.txt (or at a subpath such as /docs/llms.txt, covering the pages under it) that tells an AI agent what a site or project is and where its useful pages are. The llmstxt.org proposal (version 2, August 2026) fixes the order: one H1 with the name, a blockquote summary, details without headings, then H2 sections that each hold a list of [name](url): notes links. The page checks all of it, line by line.

It also runs the proposal's reference parser as this page reproduces it, because tools built on it are stricter than Markdown: it throws on a * bullet or a wrapped line, keeps only the last of two sections with the same name, and does not see a summary that is followed directly by the first H2. The page never fetches your site - it reads only what you paste, and it cannot tell whether a link works; a page list lets it say which links your site no longer lists. Derived from the agent skills @github/create-llms and @github/update-llms (github/awesome-copilot, MIT; see the notice).