Machine-readable capability description was incomplete
The API was described, but no separate manifest summarized selection guidance, resource-scoped pricing, rate limits, and side effects.
AI Website Intelligence
Is your website ready for AI agents?
Test whether AI systems can discover, understand, trust, and use your website. Get a scored report with evidence and prioritized fixes for agent readiness, SEO, accessibility, performance, and security.
Quick scans in seconds · Browser-rendered deep audits · Structured JSON with evidence · HMAC-signed and verifiable
Integrate the API · x402 USDC on Base · no account · from $0.002 per call
Sample data · sanitized report preview
What a Santos report looks like
Every audit returns a 0–100 AI Website Intelligence score across four dimensions, the evidence behind each finding, and fixes ranked by impact. This preview uses sanitized sample data — illustrative structure, not a customer result, testimonial, or certification.
82
Website Intelligence
91
Discoverable
78
Understandable
73
Callable
86
Trustworthy
36 of 40 checks executed · 3 not applicable · 1 skipped · 90% weighted evidence coverage · confidence 0.91. Scores are numeric labels, never color alone.
Top prioritized fixes
- Impact high · effort medium — Publish a versioned capability manifest naming each callable resource, endpoint, schemas, billing unit, price, errors, and support. Link it from HTML or an HTTP Link header.
- Impact medium · effort low — Confirm the advertised MCP endpoint in a public registry record, or remove the claim until registry evidence exists.
- Impact medium · effort medium — Add typed error responses with stable machine-readable codes to the OpenAPI document so agents can retry predictably.
Quick audits are fetch-based; Deep audits render your page in a real browser. Automated accessibility checks do not certify WCAG conformance, and passive security checks are not penetration testing.
See the output
Read a real report first
Every domain below was audited with the same engine you would be buying — AI Website Intelligence score, the four dimensions, evidence, and prioritized fixes. Read one in full, then decide. Agents pay per call: the Quick API at $0.015 USDC per success, or the Agent Readiness audit at 0.075 USDC per success.
The Santos Index · July 2026
The web isn't agent-ready — yet.
We audited 300+ of the world's best-known websites with the same engine that powers this API. Average AI Website Intelligence score: 59/100. Median: 58. google.com scores 39, oracle.com 39, binance.com 24 — while planetscale.com leads at 91. Every report is public, HMAC-signed, and independently verifiable.
Intelligence model
Four dimensions between being online and being usable
Each dimension answers a different question. Callable checks are applicability-aware, so an informational site is not penalized for not exposing an API.
Discoverable
Find the site, documentation, crawler rules, canonical URLs, sitemaps, llms.txt, and advertised interfaces.
Understandable
Interpret identity, content, services, pricing, relationships, semantic HTML, and structured data.
Callable
Use OpenAPI, MCP, capability manifests, typed schemas, stable errors, job endpoints, and x402 payment.
Trustworthy
Evaluate HTTPS, security headers, accessibility, performance, browser errors, support, and evidence quality.
Audit depth
Quick triage or browser-rendered evidence
Fetch-based · seconds
Quick
Fast fetch-and-parse signals for a single public page: timing, HTML weight, SEO, accessibility markup, security headers, and an embedded Agent Readiness view.
- Returns in seconds
- Structured 0–100 scores and fixes
- $0.015 USDC per successful paid audit
Browser-rendered · minutes
Deep
A real isolated browser adds Lighthouse lab metrics, rendered axe-core findings, screenshots, network and console evidence, and passive security analysis.
- Versioned job, report, and artifact flow
- Screenshots and rendered evidence
- $0.225 USDC per compute reservation
Both modes assess one page, not a whole-site crawl. Lab data is not field data, automated accessibility checks do not certify WCAG conformance, and passive security checks are not penetration testing.
Use cases
Built for teams making the web agent-ready
AI-agent developers
Evaluate public interfaces before adding them to a tool registry or autonomous workflow.
Agencies
Turn a URL into traceable technical evidence and a prioritized implementation backlog.
SaaS teams
Check whether product docs, APIs, payment rules, and identity claims stay consistent.
Technical SEO teams
Extend crawl and content foundations into structured meaning and machine capabilities.
Automation platforms
Consume versioned JSON without opening accounts or distributing traditional API keys.
Developer integration
Public contracts, predictable payment, no account
11 paid capabilities, metered per call in USDC on Base mainnet over the x402 protocol. No accounts, no API keys — payment settles only on a successful response, so a failed call is never charged.
OpenAPI 3.1Typed operations, schemas, errors, and x402 behaviorMCPStreamable HTTP discovery and paid audit handoffCapability manifestSelection guidance, pricing, limits, and supportllms.txtLow-noise service guide and canonical machine linksMCP serverOne endpoint, seven tools — Claude, Grok, or any MCP client
# Agent Readiness · synchronous
GET /api/agent-readiness?url=example.com&depth=quick
← 402 · PAYMENT-REQUIRED · 0.075 USDC · eip155:8453
→ retry with PAYMENT-SIGNATURE (any x402 v2 client can automate this)
← 200 · versioned evidence + PAYMENT-RESPONSE receipt
GET /api/agent-readiness?url=example.com&depth=quick
← 402 · PAYMENT-REQUIRED · 0.075 USDC · eip155:8453
→ retry with PAYMENT-SIGNATURE (any x402 v2 client can automate this)
← 200 · versioned evidence + PAYMENT-RESPONSE receipt
Human reports
Prefer a formatted report by email?
The API returns raw machine-readable JSON. Human reports are the same evidence, formatted, emailed as a private tokened link, and independently verifiable — purchased by card, no account.
One-time
Quick Report — $9
Fetch-based agent readiness report — formatted, emailed, and verifiable. Delivered same day.
One-time
Deep Report — $29
Browser-rendered website intelligence report — real-browser evidence, screenshots, formatted, emailed, and verifiable.
Subscription
Monitoring — $9/mo
Weekly re-audit of any URL, an alert when your score moves 5+ points, and a monthly digest.
About the builder
Built and operated by one accountable engineer
Santos Automation is founded and run by Juan Santos. He spent years as a hands-on field technician and in sales before teaching himself software engineering — since 2020 he has shipped more than 100 public projects, from e-commerce with live Stripe checkout to AI audit systems and automation tooling. He now runs Santos Automation full-time from Charlotte, North Carolina.
This API is production infrastructure, not a pitch deck: it meters real payments in USDC on Base, publishes its scoring methodology, and the same engine has audited 300+ of the world's best-known websites — every report HMAC-signed and independently verifiable.
No ticket queue, no account managers. When you contact Santos, you reach the person who wrote the code.
FAQ
Questions about AI Website Intelligence
What is AI Website Intelligence?
AI Website Intelligence measures whether a website can be discovered, understood, trusted, and used by AI systems and autonomous agents.
What is an Agent Readiness audit?
It checks the public technical signals that help agents find information, interpret structured meaning, and invoke applicable machine-readable capabilities.
Is Agent Readiness the same as SEO?
No. SEO mainly addresses search discovery and ranking. Agent Readiness also examines structured meaning, callable interfaces, machine-readable documentation, and operational trust.
Does llms.txt guarantee visibility in AI answers?
No. llms.txt can orient an agent and point to useful documentation, but it cannot guarantee inclusion or ranking in an AI-generated answer.
Does every website need MCP or OpenAPI?
No. They matter when a site exposes callable services. Santos marks non-applicable checks instead of penalizing an informational website.
What is the difference between Quick and Deep audits?
Quick audits use bounded fetch-and-parse checks. Deep audits use a real browser for rendered accessibility, lab performance, network, console, screenshot, and passive security evidence.