--- title: "Agent access: the machine-readable surface of hyperize.ai" description: "hyperize.ai serves every page in three forms (HTML, Markdown, JSON) and registers four WebMCP site tools in-browser agents can call: DAX 40 score lookup, insights search, an instant agent-readiness probe for any URL, and the free Snapshot request. The English homepage adds a six-tool personalization surface: a visiting agent can rewrite the page for its user. Documented here, no key, no registration." canonical: https://www.hyperize.ai/en/agents lang: en last-updated: 2026-09-01 --- # Built to be read by machines, too. Agent access hyperize.ai serves every page in three forms: HTML for people, Markdown for language models, and JSON for programs. And it goes one step further: the site registers WebMCP tools, so an in-browser agent does not have to scrape this site. It can ask for a DAX 40 score, probe a website's agent readiness, or order the free Snapshot directly. This page documents the whole machine-readable surface. Everything listed here is live. No key, no registration. Read surfaces ## One site, three formats. The same content boundary in every format. What is withheld in one is withheld in all. [llms.txt](/llms.txt) Machine-readable site guide following the llms.txt convention: what Hyperize is, when to use it, and every indexable page with a one-line summary. Scoped deep indexes at /en/answers/llms.txt, /en/insights/llms.txt, /en/dax40-index/llms.txt; everything in one file at /llms-full.txt. [Markdown twins](/index.md) Every content page has a Markdown twin: append .md to the URL, or request the page with the header Accept: text/markdown. Twins carry frontmatter (title, description, canonical) and open with the page's H1. [JSON siblings](/en/dax40-index.json) Structured JSON at the same URL with .json appended: the DAX 40 Index inventory, every scored brand page, the Insights hub (16 articles), Methodology pages, and Answer Pages. Format hyperize/v1, same content boundary as the HTML. [Sitemap with real change dates](/sitemap-index.xml) XML sitemap of every indexable URL. Content pages carry a lastmod taken from the page's own dateModified; the build date is never used as a stand-in. [Crawler policy](/robots.txt) Explicit allowlist for search and AI assistant crawlers (GPTBot, ClaudeBot, PerplexityBot, Google-Extended and others). Being read and cited by assistants is the point of this site. [Structured data](/en/methodology/agent-surface) Every page emits schema.org JSON-LD in the initial HTML: one canonical Organization entity, DefinedTerm pages for the Hyperize vocabulary (Agent Surface, Agent Success Score, Third-Party Interception), Service, Dataset and Article where applicable. [Honest 404](/llms.txt) Unknown paths return HTTP 404 with recovery links, and a Markdown body when requested with Accept: text/markdown. No soft redirects, no error pages pretending to be content. Site tools ## Four tools an agent can call. One of them acts. Every page registers these tools on `document.modelContext` (the WebMCP draft standard). Agents that support it, like the ChatGPT desktop browser, call them instead of guessing through the UI. Humans see this same page; agents get the same three things a visitor wants: check a score, read the research, order the report. get\_brand\_agent\_success\_score READ Look up any DAX 40 brand in the Agent Success Index: measured score, both axes (AI Visibility, AI Usability), the measured bottleneck, and the link to the full evidence page. Try:  “How does adidas score in the DAX 40 Agent Success Index, and what blocks it?” search\_insights READ Search the published field research on how AI agents evaluate, recommend, and transact with enterprise brands. Try:  “Find Hyperize research about agents failing at checkouts.” check\_agent\_readiness READ Instant five-check surface probe for ANY website: front door reachable, robots.txt policy for AI crawlers, llms.txt, schema.org JSON-LD, Markdown twin. Result in seconds, with DAX 40 context. A surface probe, deliberately not the measured Agent Success Score. Try:  “Check example.com: can AI agents work with it?” request\_snapshot WRITE The one write action on this site: order the free Agent Success Snapshot. Hyperize runs real agent tasks against the website, scores it, benchmarks it on the DAX 40, and delivers the report by email within 48 hours. Free, no commitment. Agents are expected to confirm URL and email with their user first. Try:  “Order the free Hyperize snapshot for our website, report to my work email.” Consequential actions stay confirmed: the one write tool submits the same form humans use, and agents are expected to confirm URL and email with their user first. Plus six on the homepage The [English homepage](/en) registers a second surface: a personalization contract that lets a visiting agent rewrite the page for its user. The agent reads per-field length budgets, submits every text in one call, and the page stages the reveal. Changes live in that visitor's browser only; a reload restores the original. get\_adaptation\_capabilities · start\_personalization · write\_all · write\_section · show\_decision\_brief · reset\_experience Try it ## From a terminal. ``` # The site guide, with a when-to-use section curl https://www.hyperize.ai/llms.txt # Any page as Markdown, two equivalent ways curl https://www.hyperize.ai/en/pricing.md curl -H "Accept: text/markdown" https://www.hyperize.ai/en/pricing # The DAX 40 Index as JSON, then one brand's sibling curl https://www.hyperize.ai/en/dax40-index.json curl https://www.hyperize.ai/en/dax40-index/brands/adidas.json # The agent-readiness surface probe (the WebMCP tool's backend) curl "https://www.hyperize.ai/.netlify/functions/agent-readiness-probe?url=example.com" ``` There is no separate API beyond these read-only surfaces and the documented tools, and no key or registration. If you are building something that needs more: [hello@hyperize.ai](mailto:hello@hyperize.ai). --- Canonical: https://www.hyperize.ai/en/agents Machine-readable index: https://www.hyperize.ai/llms.txt