LLMS-Explorer
The llms.txt standard, in practice
llms.txt tells AI agents what a site contains; LLMS-Explorer grades the files already published and maps every concept behind them in one browsable tree.
The concept tree is the map; llms files are the territory; the site keeps the two in agreement.
Get the skills and CLIBrowse the concept tree
499 concepts across 86 roots, mapped 2026-08-30· 92 of 144 scored llms files earn an A, linted 2026-09-08· 11 retrieval questions, recorded 2026-08-31
- Publishing your first llms.txtRead the rubric a file is graded against.
- Already publish oneSee how the scored files rank under the same lint.
- Here to search the corpusWatch keyword, vector and hybrid retrieval, recorded.
Explore the corpus
Tree
Every researched concept, with a 3D map of the same graph.
llms.txt and LLM-readable documentation — 5 of 5 children the same branch as mermaid
mindmap root((llms.txt and LLM-readable documentation)) llms.txt specification v2 llms-full.txt page grammars llms.txt generation tooling llms.txt ecosystem evidence llms.txt recreation and family aggregationDirectory
The mirrored
llms-full.txtfiles we can score, graded byllms_lint.Grade A, largest first — sites doing it well Demo
Keyword, vector and hybrid retrieval over one docset, recorded.
Hybrid, top hit “Install Claude Code on Windows with PowerShell”
`Claude Code on Windows requires either Git for Windows (for bash) or PowerShell`: [Install a shell](#claude-code-on-windows-requires-either-git-for-windows-for-bash-or-powershell)
Read the reference
Reference
The
/ldorubric: attributes, passes, and spokes.Opens with Formatting: the grammars side by side — The index, the three full-file grammars, the facts line, the small file, the vocabulary line, the split root and the manifest — on one page.
Examples
The decision table and copy-only recipes.
Opens with Which layer answers which question — The decision table for the cookbook: match the shape of your question to the cheapest llms layer that answers it, then open the recipe.
Blog
Launch posts with figures from the corpus.
Most recent Markdown as the LLM Output Gateway — A technical review testing the thesis that Markdown is the default machine-to-human interface layer for LLM output, covering CommonMark/GFM, streaming, and its lossiness.
llms.txt
This site's own llms family — the index an agent reads, rendered for a person.
Five files, one family - llms.txt — the index
- llms-full.txt — every page, in order
- llms-small.txt — the reference layers, budgeted
- llms-facts.txt — one claim per line
- llms-vocabulary.txt — the terms
Apply it to your own files
- Skills — the skills the tree above is built with, one page each; also in the nav menu.
- Downloads — the skills via
npx, thellmsxlibrary and CLI, the npm package, and the terminal browser. - Playground — three skills you can run in the browser: notes → llms.txt, the concept library, and the llms.txt optimizer (needs an API key with
runscope). - Contribute — add a subject to the directory. Sign in to submit, view your proposals, and track their moderation status.
- Sign in — your account, its API keys and its usage; everything above is readable without one.