How to make your business visible to AI agents (MCP, llms.txt, and schema)

AI agents increasingly act for the user — shortlisting vendors, comparing prices, even booking. To be visible to them you need three layers: machine-readable facts (structured data), a machine-readable site summary (llms.txt), and — the emerging edge — a machine-usable interface (an MCP server). This post explains each layer and what to implement first.

Layer 1: structured data (table stakes)

Agents parse JSON-LD before they parse prose. Minimum set:

• Organization — who you are, logo, sameAs profiles.
• Product / Service with Offer — what you sell and the price range. Agents filter by price; unpriced offerings get skipped.
• FAQPage — your money questions, answered.
• LocalBusiness — hours, service area, phone for local trades.

Layer 2: llms.txt (the AI crawler's front door)

llms.txt is a plain-markdown file at your site root that tells AI systems what your site is, what you offer, and which pages matter. Think robots.txt, but describing meaning instead of permissions. It takes an hour to write and many AI crawlers now request it.

Keep it under a page: what you do, who it's for, pricing model, key URLs, contact.

Layer 3: an MCP server (the emerging edge)

The Model Context Protocol (MCP) is the standard that lets AI assistants call real tools. A business MCP server exposes actions — "check availability", "get a quote", "list services" — that agents can call directly instead of scraping your site.

Why it matters early:

• Agents prefer structured calls — a tool call returns exact, current data; scraping guesses.
• Registries are young — listing your server in MCP registries now is like getting into early app stores.
• It compounds — every assistant that adopts MCP can use your server the day it ships.

For most small businesses this was out of reach — you don't build protocol servers. This is exactly the kind of plumbing TalkLytics automates: every client gets a hosted MCP server that answers agent questions about their business from their approved brand memory, listed under their name.

What to do this quarter

• Week 1: add Organization + FAQPage schema, write llms.txt.
• Week 2–4: publish answer-first pages for your top 10 money questions.
• Month 2: stand up an MCP presence and get listed in registries.
• Monthly: re-ask the four engines your money questions and track citations.

The goal isn't to chase every acronym — it's to be the easiest correct answer for a machine to give about your category.