Anthropic MCP Goes GA: Agent Commerce Infrastructure Is Now Real
Anthropic's Model Context Protocol hit general availability with a wave of enterprise integrations, making structured business exposure to Claude-based agents a production-level concern — not a sandbox experiment. Simultaneously, llms.txt tooling and AI crawler policy changes are forcing businesses to make active choices about machine-readable presence.
This Week's Signals
Anthropic MCP Reaches General Availability With Stripe, Shopify, and Atlassian Integrations
Why it matters for your score
MCP is no longer experimental — Claude agents can now query your inventory, book appointments, and process payments through MCP-connected endpoints in production environments. Businesses without an MCP server are invisible to this entire class of agentic transaction. If your competitors have published MCP endpoints and you haven't, Claude-based agents will route commerce to them by default.
Perplexity Confirms Structured Schema and llms.txt Signal Weighting in Citation Ranking
Why it matters for your score
Perplexity's engineering team confirmed that pages with valid JSON-LD business schema and a present llms.txt file receive measurable ranking preference in its citation selection layer. This is the first major AI answer engine to publicly confirm schema as a ranking input — not just a crawl aid. For local businesses and service providers, this makes structured data directly tied to whether you appear in an AI-generated answer at all.
GPTBot Expands Crawl Scope to Include Pricing Pages and FAQ Structured Data
Why it matters for your score
OpenAI updated GPTBot's crawl directives to specifically prioritize pages with PriceSpecification and FAQPage schema markup, signaling that ChatGPT's business recommendation layer is actively being trained on structured pricing and service data. Businesses with unstructured or missing pricing pages are handing that training signal to competitors who have marked up theirs. This also means blocking GPTBot in robots.txt now has a direct cost: your pricing data won't feed ChatGPT's recommendation model.
What to do this week
- 1
Publish or update your llms.txt file immediately — Perplexity's confirmed weighting means this is now a citation ranking factor, not optional hygiene. Use Faro's /tools/llms-txt generator to produce a validated, machine-readable business summary and deploy it to your root domain this week.
- 2
Add PriceSpecification and FAQPage JSON-LD schema to your top service and pricing pages before GPTBot's next crawl cycle. Run Faro's /tools/ai-schema tool to generate compliant markup, then verify GPTBot is permitted in your /tools/robots-txt settings — blocking it now costs you training signal in ChatGPT's recommendation layer.
- 3
Assess your MCP exposure gap using Faro's /tools/ai-readiness-scan — the scan now flags whether your domain has a discoverable MCP endpoint. If you're in e-commerce, professional services, or SaaS, MCP GA means agent-driven transactions are live; a zero score here means you're invisible to Claude agents routing real commerce today.