AI is changing how the world buys.If AI can’t read your site, you don’t exist.
When your customer asks ChatGPT, Claude, or Perplexity about your category, the answer comes from the data your site exposes. The Agent Kit ships the three files AI reads first — agents.json, llms.txt, agent-instructions.md — so the answer includes you. $49 once, two minutes to deploy.
~2 min delivery·No code to write·$49 once·No subscription
Works with Shopify, WordPress, Webflow, Wix, Squarespace, and any custom website. See the Lighthouse audit →
yourstore.com — no agent files
Your customer asks AI where to buy.
ChatGPT, Claude, and Perplexity answer “where should I book?” and “what should I buy?” millions of times a day. The answer is assembled from the sites the AI can actually read.
The agent opens your site — and reads noise.
No agents.json, no llms.txt. It guesses your navigation, misreads your forms, burns its token budget on markup that was written for human eyes — and gives up.
You're skipped. Silently.
There’s no log line, no abandoned cart, no error email. The recommendation simply names someone else. Most owners only see it later, in the year-over-year revenue chart.
Agents optimize for one thing.
Highest task-success at the lowest token cost. Every page an agent has to parse costs tokens; every guess costs retries. Structure is cheap. Guessing is expensive.
Three files make you the cheapest answer.
agents.json— the action map: every form, flow, and endpoint an agent can use, with parameters and success indicators.llms.txt— the condensed brief: who you are and what you sell, written for a small context window.agent-instructions.md— the runbook: how to behave on your site, where the gotchas are, what to prefer over scraping.
Cheap to resolve gets recommended.
That’s the Law of Least Token Resistance: being named in AI answers isn’t a favor — it’s a side effect of token economics. Drop the files in your site root; agents discover them the same way they discover robots.txt.
Whatever you sell, the mechanism is the same.
The agent reads
- products
- 142 items, priced
- shipping
- EU + US, 2–4 days
- returns
- 30-day policy
✓ add_to_cart — mapped
Your catalog, ingested in one pass.
Average agent-readiness score, before and after the kit
Average agent-readiness score across 11 real sites, before and after the kit. Every site lifted — minimum +37, maximum +74. Read the full study →
Don’t take our word for it. Watch an AI.
Two free checks. Enter your store and see exactly what an AI can — and can’t — do with it right now.
Free. No email, no signup. Works on any store.
Each check runs a live AI read of your store — honest, specific, and something you can re-run in ChatGPT yourself.
What the kit closes,
and what it leaves to your CMS.
Chrome’s Lighthouse Agentic Browsing category scores nine specific audits. Flip the toggle to see what happens when the kit ships.
llms-txt-presentKit closesIs /llms.txt reachable with plain-text content type?
File-layer
llms-txt-well-formedKit closesDoes the file parse against the llmstxt.org reference?
File-layer
agents-json-presentKit closesIs /agents.json reachable as valid JSON?
File-layer
agents-json-actions-typedKit closesDoes every action have typed parameters (no null)?
File-layer
agent-runbook-presentKit closesIs /agent-instructions.md reachable?
File-layer
auto-discovery-linksKit closesThree <link rel="alternate"> tags in homepage <head>?
File-layer
sitemap-discoverableCMS-sideIs /sitemap.xml valid and referenced from /robots.txt?
Your platform
schema-org-densityCMS-sideSchema.org JSON-LD on homepage and primary pages?
Your platform
webmcp-annotationsSpec still movingWebMCP per-element annotations on interactive elements?
Nobody passes in 2026
0 of 9 audits closed — site ships no kit files
The remaining three are honest scope-bounding — `sitemap-discoverable` is a one-line CMS config edit, `schema-org-density` is theme-side and walked through in each platform pillar, and `webmcp-annotations` is a spec that hasn’t settled yet (nobody passes it in 2026).
From URL to kit, in minutes.
No setup, no SDK, no integration work. One URL, one payment, one zip. Standards-aligned output that won't break when the agentic web standardizes around it.
- 01
Submit your URL
We crawl the surface area of your site with Firecrawl — pages, schema, forms, endpoints.
- 02
Claude maps the actions
A deep action audit identifies every intent an agent could perform, with locators and parameters.
- 03
Download your kit
Three files in a zip — drop them in your site root and you're agent-ready.
Agents are arriving. Most sites are unreadable to them — yours doesn’t have to be.
We don’t touch your stack and we don’t inject code. The kit is three plain text files served from your domain, the same way the entire web serves robots.txt.
100% passive static files
≈ 12 KB totalJSON + Markdown in your site root — the robots.txt mechanism.
Zero code execution
No PHP · No JSNo plugin, no script, no database. Nothing to attack, nothing to slow down.
Zero performance impact
CWV-safeCached at the edge. Lighthouse scores don't budge by a point.
Server-side, no scrape leaks
Nothing storedYour URL is processed server-side; the scrape is discarded once the kit is built.
How fast does the kit
pay itself back?
Plug in your average customer value and your current monthly volume. The kit is $49 once — most stores recover it on a single transaction.
Your numbers
Or try a preset
Your payback
$49 kit recovered after
1 customer
At your current rate, that’s today.
Kit price
$49 once
Avg customer
$49
Math assumes one additional recovered customer attributable to AI-search visibility. The actual lift compounds as agent traffic grows quarter over quarter.
See what AI reads on your site — right now.
Free scan, no email. If the score is low, the kit fixes it for $49 — once.