# llms.txt

> A markdown file at the domain root that gives language models a curated map of a site's most important content.

_The Agent Protocol Atlas · /protocols/llms-txt · [JSON](/api/protocols/llms-txt) · [all The Agent Protocol Atlas](/protocols)_

- **name:** llms.txt
- **full_name:** llms.txt
- **layer:** discovery
- **creator:** Jeremy Howard (Answer.AI)
- **status:** widely adopted (~10% of studied sites)
- **year:** 2024
- **one_liner:** A markdown file at the domain root that gives language models a curated map of a site's most important content.
- **spec_url:** https://llmstxt.org
- **snippet:**

```
# Site Name
> One-line description.
## Core
- [Page](/page.md): what it is
```
- **abbreviation:** llms.txt
- **also_known_as:** /llms.txt, llms.txt standard
- **canonical_spec_url:** https://llmstxt.org
- **entity_uri:** https://llmstxt.org
- **taxonomy_layer:** discovery
- **sub_layer:** site-summary-file
- **protocol_type:** declaration
- **central_problem:** Gives language models a curated, markdown map of a site's most important content so they don't have to parse complex HTML, ads, or JavaScript.
- **maintainer:** Jeremy Howard / Answer.AI + community (llmstxt.org)
- **governance_body:** none (community proposal)
- **license:** Apache-2.0
- **maturity_tag:** emerging
- **current_spec_version:** — (verify-against-primary-at-build)
- **spec_date:** — (verify-against-primary-at-build)
- **launch_date:** 2024-09-03
- **last_verified:** 2026-06-15
- **transport:** Static markdown file at the domain root
- **core_mechanism:** A markdown file at /llms.txt uses a fixed shape — H1 site name, a blockquote summary, then sectioned lists of links with short descriptions — so an LLM gets a concise, parseable map of the site's key pages (often with .md twins).
- **discovery_endpoint:** /llms.txt (and the fuller /llms-full.txt)
- **settlement_type:** —
- **adoption_metric:** Approximately 10% of studied sites by May 2026; adopters include Anthropic, Stripe, Cloudflare, Vercel, Mintlify, Supabase (source: https://www.answer.ai/posts/2024-09-03-llmstxt.html)
- **notable_adopters:** {"value":"Anthropic","source":"https://codersera.com/blog/llms-txt-complete-guide-2026/"}, {"value":"Stripe","source":"https://codersera.com/blog/llms-txt-complete-guide-2026/"}
- **relationships:** {"predicate":"complements","target":"agents-json","note":"llms.txt declares content; agents.json declares APIs/workflows — paired Layer-1 discovery files."}
- **ideal_use_case:** Helping LLMs find and quote the right pages on a content/documentation site.
- **when_to_use:** When you want models to read and cite your most important content — the cheapest, first agent-readiness step.
- **when_not_to_use:** When agents need to take actions or call functions (use MCP/WebMCP/agents.json), not just read curated links.
- **code_example:** # Acme Docs
> Acme's developer documentation.

## Core
- [Quickstart](/quickstart.md): get started in 5 minutes
- [API Reference](/api.md): every endpoint
- **source:** Origin proposal (Jeremy Howard, Answer.AI), 2024-09-03: https://www.answer.ai/posts/2024-09-03-llmstxt.html . Spec home: https://llmstxt.org .
- **agent_readiness_link:** agent-readiness/llms-txt
- **layer_legacy:** content