Concierge Beacon
Your website, readable and callable by AI agents. A public MCP server, published Agent Skills, llms.txt, and in-page WebMCP tools — so when a customer's assistant asks about your product, the answer comes from you.
What you get
Four discovery surfaces on your existing site, generated from one content source. A hosted MCP server that exposes your pages, pricing, and product facts as callable tools. Agent Skills published under /.well-known/ so capable assistants load your procedures directly. An llms.txt index that tells AI crawlers what your site is and where the canonical answers live. And WebMCP tools registered in the page itself, so browser-resident agents act through a sanctioned surface instead of scraping your DOM.
What makes it different
Most sites are invisible to agents or misread by them. Beacon generates all four surfaces from the same content source in your build pipeline, so a copy change can never leave an agent reading a stale version of your product. This site runs the same setup — groupemedia.ca publishes its llms.txt and registers WebMCP tools — and so does the production platform we operate.
What it can do for you
Buyers increasingly ask their assistant before they ask a search engine. Beacon makes sure the assistant's answer about your product comes from your content, current and complete.
- Public MCP server hosted at the edge, with rate limits
- Agent Skills index generated from your real content
- llms.txt and llms-full.txt kept in sync with the site
- In-page WebMCP tools: search, pricing, contact, forms
- Regenerated on every deploy — no stale agent answers
- Analytics on which agents visit and what they ask
When it makes sense
Your buyers use ChatGPT, Claude, or an AI-powered browser as a research step. Your product facts change often enough that scraped snapshots go wrong. Or you sell to technical teams who will judge you on whether your own site is agent-accessible.
How we ship it
Typical deployment is 1–2 weeks on top of your existing site — no rebuild, no CMS migration. We wire the generation into your build pipeline, so the surfaces stay truthful without anyone maintaining them. Everything runs under your domain and your infrastructure.
Questions people actually ask
Why should we care whether AI assistants can read our site?
Because your buyers increasingly ask ChatGPT or Claude before they ask Google. Without Beacon, those assistants answer from scraped snapshots that may be stale or wrong. With it, the answer about your product comes from your own content, current and complete — pricing, facts, and procedures included.
Does this expose anything private?
No. Beacon only publishes what's already on your public site — pages, pricing, product facts — through sanctioned surfaces: a hosted MCP server (a standard connection AI tools use to call your content), Agent Skills, llms.txt, and in-page WebMCP tools. Everything runs under your domain with rate limits, and you get analytics on which agents visit and what they ask.
Do we need to rebuild our website?
No rebuild, no CMS migration. A typical deployment is 1 to 2 weeks on top of your existing site: we wire the generation into your build pipeline so all four surfaces regenerate on every deploy. Nobody has to maintain them by hand, and a copy change can never leave an agent reading a stale version.
What if we later want agents to read more than our public site?
Beacon covers the public side; when you want your private corpus — internal docs, runbooks, customer notes — exposed the same way with permissions and audit, that's Library MCP, on the same stack. Everything Beacon generates lives in your build pipeline and under your domain, so it stays yours either way.
Frequently paired with
Related products
An AI concierge for your website, app, or product — knows your brand, knows your data, knows your tools, and routes every conversation to a real outcome.
Your private knowledge corpus, exposed as a single MCP server — so Cursor, Claude Desktop, ChatGPT Desktop, and any MCP-aware AI client can read it directly, with governance and audit applied.