Agent to agent

MCP for your brand.

Your website was built for people who look. Agents do not look, they ask. Parleon gives your brand one Model Context Protocol server and points it in two directions: outward at the assistants your customers use, and inward at your own team.

External MCP

So the world can buy from you.

Your customers, their agents, and the assistants they already use

search_catalog

A customer asks ChatGPT for someone who can do the job this week under a set budget, and what you actually offer comes back, with the detail, the price and the nearest location.
get_quote

The agent puts a real figure on the table, assembled from your own rates, with its provenance and its caveats attached to the number.
get_offers

Your packages and current promotions, live, so nobody is quoted an offer that ended last Tuesday.
book_appointment

When they are ready, the slot is taken on your calendar, or the conversation lands in your CRM as a structured enquiry with the whole context intact.

Internal MCP

So your team stops going to look.

Your own team, your marketing agency, and whatever AI tools they already work in

get_capacity

Live open capacity by location and by service, with everything attached: price, duration, next available date. The question a marketer actually starts with.
get_offers

Every current package and promotion across the group, in one answer, so creative is built on what is live today.
get_pricing

What every location is actually advertising right now, without opening a single site.
campaign_status

What is running, what is paused, what changed since Friday. Your agency asks the brand directly instead of asking you.

Same engine, same data, same brand brain. The only difference is which direction it faces and who is allowed to ask.

Monday morning, planning the month

Nobody opens twenty websites to count openings any more.

The marketing lead is in ChatGPT. ChatGPT is talking to the brand. Every number below came back from a live MCP call.

Which services should we advertise this month?

Asked your brand for live capacity get_capacity

Open capacity next 30 days, by service

Routine service visit	412
Annual inspection	388
Emergency callout	240
Install and fit-out	197
System upgrade	142
Warranty follow-up	96

1,475 open appointments across 20 locations, counted at the last sync with nothing dropped since the one before it.

Lead with routine service visits and annual inspections. You have 412 and 388 slots open in the next thirty days, so the spend points at work you can actually take. Emergency callout is thinner at 240 and it fills itself anyway, so treat that one as secondary.

Before, that answer meant opening every location's site, or waiting on someone who would.

Your knowledge, not just your catalog

AI would rather look it up than guess.

The catalog is the easy part. It is already a list. The hard part is everything else your business knows and never wrote down in one place: how financing actually works at your store, what the service menu covers, which trim comes with what, the warranty terms, the answer your best salesperson gives that lives only in their head.

We take that material in whatever shape it arrives, your website, documents, PDFs, spreadsheets, and turn it into something an assistant can read and answer from. Not a search box that hands back ten links. An answer, with its source attached.

This is the part people call RAG. The plain version: an AI is far more accurate when it can look something up than when it is working from memory. So we build the thing it looks in, and we keep it current.

It faces both ways, like everything else here. Outward, a shopper's assistant answers a finance question correctly at eleven at night. Inward, your team and your agency ask the same question and get the same answer, instead of three people guessing differently.

One group's knowledge base today

468

documents, built from 43 pages of their own site and organised the way questions actually get asked.

Individual models: 308
Departments and locations: 54
How it all relates: 51
Straight answers to common questions: 42
Guides: 12

Every document records the page it came from, so an answer can cite where it got that from rather than asking anyone to take its word. 1,828 documents across the fleet.

Plugins, extensions, custom MCPs

A hub for your brand, so one agent can reach everything.

Every tool you use is growing its own MCP server. Left alone, that is a dozen disconnected doors and an agent that can only open one at a time. Connect them to Parleon and your brand becomes the single place worth asking, because the answer can finally cross the tools.

Meta Ads

Campaigns, spend, creative performance
Google Ads

Search and PMax campaign state
Your CRM

Leads, ownership, follow-up status
Analytics

Sessions, conversions, sources

Catalogue feed

Live services, products, pricing, photos
Reputation

Reviews and response state
Scheduler

Staff, availability, bookings
Your own MCP

Anything you already run

One question, many systems

"Which campaigns are promoting something we can no longer take on?" needs the ad platform and the capacity feed in the same breath. Federated, that is one call.

Add a tool, not an integration

A new platform ships an MCP server, you register it once, and every agent already talking to your brand can use it immediately.

Your rules at the door

Scope stays yours. Customers reach what is public, your team reaches margin and campaign state, and nothing crosses that line by accident.

The internal case

Twenty locations is twenty tabs. Or it is one question.

Right now, finding out whether a special expired on one of your sites means someone opening that site. Multiply by every location, every week, and it becomes a job nobody wants and everybody skips. The reason it is manual is that your data has never been callable. It is now.

Ask the portfolio, not the pages

"Which of our stores still shows a July special?" comes back as a list, not as an afternoon.

Give your agency a source

They build from live availability, live prices and live offers, so an ad cannot promote something you stopped offering on Thursday.

Bring your other tools in

Why a protocol and not an API

An API waits to be integrated. A protocol is already spoken.

Nobody is going to build a custom integration with your brand. But every serious agent already speaks MCP, so the work of being reachable happens once, and every new assistant that appears can talk to you on day one.

It is the difference between being listed somewhere and being callable. One is a directory entry. The other is your brand answering the phone.

One brain, two doors: External and internal read the same model. Fix a price once and both the shopper's agent and your team see it.
Scoped access: Customers get what is public. Your team gets margin, aging, campaign state and the questions agents could not answer.
Data that dates itself: Every answer carries when it was last verified, so an agent knows whether to trust it and your team knows what to chase.

Point it out, point it in, and connect the rest.

We stand up your brand's MCP server, wire it to your live data, give your team and your agency internal access, and register the other MCP servers you already run.

Free, about a minute, no account. Or book a demo.