Faro AI Signals·July 6, 2026

Anthropic MCP Goes GA: The Agent Handshake Is Now Standard

Anthropic's Model Context Protocol reached general availability with broad platform adoption this spring, making structured agent-facing endpoints a baseline expectation rather than an experiment. Businesses without MCP-compatible surfaces are now invisible to an expanding class of autonomous agents executing real transactions.

ShareShare

This Week's Signals

1

MCP General Availability Triggers Wave of Platform Integrations Across SaaS and E-Commerce

Source: Anthropic

Why it matters for your score

With MCP now GA and adopted by Shopify, Cloudflare, and dozens of SaaS platforms, AI agents running on Claude — and increasingly on other models via the open spec — can discover, authenticate, and transact with businesses that expose MCP-compatible endpoints. If your business runs on a platform that has shipped MCP support, your catalog, pricing, and availability data may already be agent-readable — but only if your schema and permissions are correctly configured. Businesses not yet on an MCP-enabled platform need to treat this as an infrastructure gap, not a future consideration.

2

Perplexity Tightens Local and Vertical Citation Logic, Favoring Structured Business Profiles

Source: Perplexity AI

Why it matters for your score

Perplexity's answer engine has shifted toward citing businesses with explicit structured data signals — particularly JSON-LD LocalBusiness, Service, and FAQPage markup — over plain-text directory listings when surfacing recommendations in shopping and professional services queries. Businesses with clean schema implementations are appearing in cited answer cards; those relying on third-party profile pages without owned structured data are being displaced. This is the clearest confirmation yet that on-site schema directly influences AI citation frequency, not just traditional search ranking.

3

llms.txt Adoption Crosses 40,000 Indexed Domains as Crawlers Begin Active Parsing

Source: llmstxt.org

Why it matters for your score

The llms.txt standard, proposed by Answer.AI's Jeremy Howard, has crossed a meaningful adoption threshold with verified parsing now confirmed from ClaudeBot and at least one Perplexity crawler variant. The file gives AI models a machine-readable summary of who you are, what you offer, and which URLs matter — cutting through the noise of a full site crawl. Businesses that have not yet published an llms.txt file are surrendering a direct, low-cost channel to shape how AI systems summarize and recommend them, especially for long-tail and conversational queries where crawl depth is limited.

What to do this week

  1. 1

    Publish or audit your llms.txt file this week — ClaudeBot and Perplexity crawler variants are confirmed parsing them. Use Faro's /tools/llms-txt generator to build a structured file that includes your core service descriptions, pricing page URL, and a plain-language business summary. Treat it like your AI business card: if it's missing or stale, you're under-represented in model context.

  2. 2

    Run a Faro AI Readiness Scan (/tools/ai-readiness-scan) specifically against your JSON-LD schema to check for LocalBusiness, Service, and FAQPage completeness — Perplexity's updated citation logic is directly rewarding these signals in answer cards. If your schema is thin or missing the 'priceRange', 'areaServed', or 'hasOfferCatalog' properties, use /tools/ai-schema to patch those fields before next week's crawl cycle.

  3. 3

    If your platform (Shopify, any Cloudflare Workers-based stack, or a major SaaS CRM) has shipped MCP support, verify that your product catalog, availability, and pricing endpoints are correctly permissioned for agent access — check your robots.txt isn't blocking MCP-relevant paths using Faro's /tools/robots-txt auditor. If you're on a platform without MCP support yet, flag it as a vendor evaluation criterion for Q3; agent transaction traffic is no longer hypothetical.

Sources

← All editionsScan your site now →