Agent readability for any website.
We make your site readable to AI agents.
When someone asks an assistant what to buy or who to call, the answer is built from sites it could actually read. Yours hands it the theme, the trackers, and the cookie banner. One install serves your facts as clean markdown instead, records every agent visit, and changes nothing for humans or Google.
Free forever. Any site, any page, no request limit.
One install. Every assistant your customers already use.
- ChatGPT
- Claude
- Gemini
- Perplexity
- Copilot
- Mistral
ChatGPT, Claude, Gemini and Perplexity never ask for clean content. We serve it to them anyway.
The proof
What an agent receives: before vs. after
An assistant cannot cite what it cannot read.
Commerce PDP capture
98.6% smaller69× less context
Before: raw PDP HTML
91,226 bytes · ~22,789 tokens
After: gateway markdown
1,315 bytes · ~328 tokens
Quotable, not just smaller. An assistant lifts these facts straight into its answer.
Why it pays
The channel is worth showing up in
Independently measured, by people who do not sell you anything.
~4.2x
content negotiation vs. llms.txt
Content negotiation (Accept: text/markdown) measured ~4.2x more effective than llms.txt for accurate retrieval (300k-domain study, mid-2026).
UCP · ACP · MCP
the commerce protocols are live
UCP launched Jan 2026 (Google + Shopify; Etsy, Wayfair, Target, Walmart backing). ACP (OpenAI + Stripe) live since Sept 2025. All three major card networks support agent-initiated payments.
Any site, any page
It works for your kind of site
Every agent wants the same thing from a page: the facts, in a fixed order, without reading the theme. A dentist’s page needs different facts than a newsroom’s, so we published the list for every kind of page.
- Place or location page
- Service page
- Collection or listing page
- Product page
- Article or post
- Reference document
- FAQ page
- Unclassified page
What each kind has to state → the facts an agent needs, per page kind, from the open vocabulary.
How it works
Install once. Nothing else changes.
Agents get your facts as clean markdown. Humans and Google see your site exactly as before.
Shopify
Paid: we run and monitor the endpointOne-click install. Nothing to configure.
How App Proxy serving works →WordPress
Free forever: it never touches our serversActivate the plugin. That is the entire setup.
What the plugin serves →Any stack
Free forever: it runs in your stackOne middleware, wired to your own source of truth.
Read the quickstart →All three install paths, in detail → with the snippet, what each one serves, and who runs it.
Your dashboard
Watch your answer rate climb
Every agent request on your site, recorded: which assistant, which URL, what it was served, and whether you had an answer. The URLs you could not answer arrive as a list, with the fix beside each one.
Visits
An agent fetches your pages through the gateway.
Answered
The gateway serves that request as markdown instead of your theme.
Referred
The visitor arrives carrying an AI platform referrer, recorded when the platform sends one.
Corroborated
That visitor landed on a page the same assistant had already fetched, which is evidence it was in the answer.
Conversions
The order or booking is attributed to agent traffic.
Every stage is counted, not modelled. A referred arrival needs the platform to send a referrer, and not all of them do. Nobody can tell you how often an assistant cited you, because the platforms hold that and do not publish it. So we show the page it fetched and the visitor who then landed there, and call that what it is.
Agent intent
See what agents asked for
Traffic tells you agents came. Intent tells you what they wanted: the searches they ran, the ones your site came back empty for, and the visitors AI conversations sent you.
What agents searched
The literal queries agents sent your site, through site search and protocol endpoints. Counted, PII-screened, never guessed.
Demand you could not answer
Zero-result searches, listed by query, so you can see what agents wanted and left without.
Visitors sent by AI
Humans arriving from AI conversations, by platform, and the orders and bookings that follow, labeled as inference.
The public record
The parts you can check without trusting us
A scanner anyone can run, an index nobody can buy into, and a standard anyone can re-implement. No account needed. No site is named without its owner’s verified opt-in.
The scanner
Paste any public URL and see what an agent actually receives from it, scored against the open Agent Readability Spec. Every check shows the evidence it was scored on.
/scan
The readable web
Aggregate readability distributions over everything we scan, published only above a floor of 25 independent domains per bucket. Named entries are verified opt-in only, and removal is free, self-serve, and permanent.
/readable
Agent Traffic Index
Which agents are reading the web, aggregated from gateway traffic. The page shows the live count toward its 25-store floor. Nothing publishes until the floor is met.
/agent-traffic
The ARS standard
The scoring spec itself: versioned, deterministic, and reproducible from its conformance corpus, with every weight and threshold published as a number rather than a description.
/spec/ars
Our scanner identifies itself on every request and obeys robots.txt. See our crawler policy, request budget, and opt-out.
Trust
Your data never feeds a negotiation engine
Per-merchant data is never sold, shared, or used to negotiate against you.
How the data wall works →Pricing
The gateway is free forever
Wire it yourself and pay nothing, forever. Pro finds what agents could not read on your site, puts it in order, and writes the fix. Enterprise is contracted.
Free
$0
forever
For teams who will wire it themselves. The gateway runs in your own stack and answers every agent request, unmetered, on every plan. What the free Console analyses is one site and 1,000 agent visits a month.
Start freePro
$99/mo
or $990/yr (two months free)
One price, whatever your traffic. All of it runs on our servers, never in your serving path. The Miss Report runs on your own visits, so it pays from day one. Pro covers up to 3 sites · every agent visit analysed · Answers Hub on 3.
Start Pro →Agency
$499/mo
or $4,990/yr (two months free)
For agencies and multi-brand operators. It sizes properties, not traffic, so the price is a number you already know before you arrive. Agency covers up to 25 sites · every agent visit analysed · Answers Hub on 3.
Start Agency →Compare every row → including which rows are shipped today and which are still planned.