Agents

How AI agents and language models read poe2k.com: every standard and format the site follows for them, where each one is, and how to find it.

<page> is a page’s own address and <section> a section’s; each is shown with the address it makes for /db/tools and /db. Every address here is on https://poe2k.com.

Read the pages

Every page reads without JavaScript, and has the same content as Markdown and its facts as JSON-LD.

Markdown of every page

A page’s content without its navigation and scripts, opening with YAML front matter: title, description, url, the section of a page that has one and, for articles, updated.

llms.txt

What poe2k is, how to describe it, and an index of the sections; each section has its own scoped llms.txt.

  • Where: /llms.txt, <section>/llms.txt (/db/llms.txt)
  • How to find it: Link: </llms.txt>; rel="describedby" on every response, and <link rel="describedby"> in every head, naming the section’s own where there is one.
  • Specification: The /llms.txt file (v2), convention

Structured data

Typed facts about each page: a breadcrumb trail, and what the page is (a collection, a person’s profile, a tool as a SoftwareApplication, an article).

  • Where: each page’s head, as an application/ld+json script
  • How to find it: In the HTML before any script runs.
  • Specification: schema.org in JSON-LD, standard

Schemamap

Each page’s JSON-LD graph on its own, and an index of them with their schema.org types.

  • Where: <page>.jsonld (/db/tools.jsonld), /schemamap.xml
  • How to find it: <link rel="alternate" type="application/ld+json"> and <link rel="schemamap"> in the head; the API catalog, under a URI relation.
  • Specification: Schemamap, proposal

Find what exists

Each of these answers “what does this site offer” from one fetch, without reading any page.

Every response, an image’s or a JSON file’s too, links llms.txt, the API catalog, the sitemap, the feeds, the skills and the AI catalog; a Markdown file’s links its section’s llms.txt.

  • Where: every response
  • How to find it: curl -sI https://poe2k.com/
  • Specification: RFC 8288: Web Linking, standard

API catalog

A linkset of every API and machine-readable resource, each with its description and documentation.

AI catalog

The Agent Skill, with its media type and example questions.

Agent Skill

A SKILL.md that teaches an agent when to use poe2k and how to read it; the index has its sha256 digest.

Page index

Every page as JSON with its addresses, section, schema.org types, headings, update and token estimate.

Sitemap

Every page search engines get, with its last change where the page knows it.

AI Discovery Document

The news API’s actions, for agents that read this draft.

Feeds and APIs

News in the formats feed readers and agents already read.

RSS and JSON Feed

Every overlay release and every post, newest first, in full.

News API

Every published article as JSON, and each one with its Markdown; read-only, no auth.

What agents may do with it

poe2k wants to be read, cited and learned from: every crawler may fetch everything, for search, live answers and training.

robots.txt, by name

Allows everyone, and names the AI crawlers (GPTBot, ClaudeBot, PerplexityBot and others), so the policy is explicit.

Content Signals

Content-Signal: search=yes, ai-input=yes, ai-train=yes in each robots.txt group.

Shared training preference

User-agent: __usage_ai-training with Allow: /: in favour of training, for every provider that reads it.

TDMRep

tdm-reservation: 0: no reservation against text and data mining.