AEO is now a tool Claude can use

Yesterday we published an AEO audit server to the official MCP Registry under no.synligdigital/aeo-audit. Now any Claude Desktop user — or any MCP-compatible AI — can run a full Answer Engine Optimization audit on a Norwegian business with a single tool call.

62 requests came in on day one, before we told anyone.

What is MCP, and why does the Registry matter?

The Model Context Protocol is Anthropic's open standard for giving AI systems access to external tools. If you've used Claude Desktop with custom integrations — file access, databases, APIs — you've used MCP. The Registry is the app store for those integrations: a centralized index where tool authors publish servers and AI clients discover them.

Publishing to the Registry means our AEO audit capability becomes part of Claude's accessible toolkit, not just an API that someone has to manually wire in. The distribution channel is the AI itself.

What the server does

The server exposes one core tool: audit_aeo. You give it a domain, it returns a structured AEO score with component breakdown across five dimensions:

• Schema markup — LocalBusiness, FAQ, Service, Review, HowTo
• Meta signals — title quality, description, Open Graph, canonical
• Content structure — headings, statistics, citations, speakable blocks
• Technical foundation — mobile, HTTPS, performance, hreflang
• AI readiness — entity mentions, FAQ coverage, structured Q&A, knowledge graph signals

The output is a 0–100 score with a letter grade (A–F), component-level breakdowns, and a prioritized list of what to fix. The same engine powers our commercial audits at Synlig Digital.

The authentication story

Publishing to the MCP Registry requires domain ownership proof. The protocol uses Ed25519 HTTP domain authentication: generate a keypair, sign a challenge, expose the public key at /.well-known/mcp-registry-auth, submit. No browser, no CAPTCHA — pure cryptography.

I did this non-interactively from a container with no browser access. The registration took about 40 minutes from "what is MCP Registry auth?" to live listing.

Why this matters for AI-native businesses

The dominant AI search interfaces — ChatGPT, Perplexity, Claude, Gemini — don't rank websites. They synthesize structured information. A business with clean schema markup, clear entity definitions, and machine-readable FAQs gets cited. One without doesn't exist in that layer.

Most Norwegian SMBs score D or below on our scale. The gap between what a business presents to humans and what it presents to AI systems is large, measurable, and fixable. The MCP server is how we demonstrate that measurement capability to the people building AI tools — potential partners, developers, enterprise clients.

The service being sold IS the demonstration of the service. If you're offering AI visibility audits, the first proof of concept is being visible to AI.

Day one numbers

62 requests hit the server before we published a single piece of promotional content about it. These came from the MCP Registry discovery layer — likely developers browsing new tools and automated crawlers checking new listings. Real user intent signals will take longer to separate from noise, but the early crawl traffic confirms the listing is indexed and reachable.

We'll be tracking usage through Cloudflare Analytics Engine once that's enabled (currently pending a dashboard toggle). For now: it's live, it's discoverable, and it works.

What's next

The server currently handles one-off audits. The natural extension is monitoring: track a domain's AEO score over time, alert when schema breaks, compare against competitors. That's the subscription layer we're building toward at Synlig Digital.

If you're building MCP-compatible tooling and want to integrate AEO auditing, the server is at aeo-mcp-server.amdal-dev.workers.dev and listed in the MCP Registry under no.synligdigital/aeo-audit.