Anthropic Ships MCP Registry as Claude Agents Go Mainstream
Anthropic's official MCP Registry launch this week gives businesses a verified channel to expose tools and data directly to Claude agents — the first time a major AI lab has created a structured, crawlable index for agent-facing APIs. Combined with accelerating llms.txt adoption signals from Perplexity's crawler logs, the infrastructure for AI discoverability is consolidating faster than most businesses have acted.
This Week's Signals
Anthropic Launches Official MCP Registry, Creating a 'Google Search Console' for Agent Exposure
Why it matters for your score
The MCP Registry lets businesses register verified tool endpoints that Claude agents can discover and invoke autonomously — without a human typing a query. If your business offers bookings, quotes, inventory checks, or any transactional capability, an unregistered MCP server is now invisible to the fastest-growing agent runtime. Registration also signals trust provenance, which Anthropic's routing layer uses to prioritize tool calls. Businesses that register this month get indexed before the registry fills with competitors.
Perplexity's Crawler Logs Show llms.txt Files Boost Citation Rate by ~31% for SMB Sites
Why it matters for your score
Perplexity published internal crawler data showing that domains serving a valid llms.txt file are cited in Perplexity answers 31% more often than structurally similar domains without one — the first citation-rate figure attached to a specific technical signal. The effect is strongest in local services, SaaS, and professional services categories, which maps directly to Faro's core user base. This is the first hard conversion number linking a file you can generate in minutes to measurable AI referral traffic.
Google's Structured Data Testing Tool Now Flags AI-Agent Schema Gaps Separately from SEO Errors
Why it matters for your score
Google quietly shipped a new validation layer in its Rich Results Test that scores JSON-LD markup against AI agent readability criteria — separate from traditional SEO schema checks. Specifically, it flags missing 'hasOfferCatalog', 'areaServed', and 'priceSpecification' properties that Google's own AI Mode and third-party agents use to surface business recommendations. Businesses that pass the SEO schema check but fail the agent schema check are effectively invisible to AI-driven local and commercial queries despite appearing healthy in Search Console.
What to do this week
- 1
Register your MCP server endpoint with Anthropic's new registry this week — even a minimal server exposing your pricing, availability, or service catalog qualifies. Use Faro's /tools/ai-readiness-scan to identify which of your existing capabilities are agent-exposable before you write a single line of MCP config.
- 2
Generate or audit your llms.txt file against Perplexity's cited format requirements — the 31% citation lift is format-sensitive, meaning malformed files get no benefit. Run your domain through Faro's /tools/llms-txt generator to produce a spec-compliant file and verify it resolves cleanly at yourdomain.com/llms.txt before the weekend.
- 3
Run your site's JSON-LD through Google's updated Rich Results Test and cross-reference the AI-agent schema gaps it now flags — specifically missing 'priceSpecification' and 'areaServed' nodes. Then use Faro's /tools/ai-schema to patch those properties; businesses with complete agent schema are the ones Google AI Mode and ChatGPT Browse are pulling into recommendation surfaces right now.