“If I asked my customers what they want, they would tell me they wanted a faster horse”

Henry Ford

Software is undergoing a revolution.  I am lucky enough to have access to people at the sharp end of this, through the engineering team at Greenhouse OS.   

What do you need to know?

• AI is now writing code at a pace that allows developers to build programs much, much faster than they used to
• Connections are now being built that allow AI systems to talk to other software programs and complete tasks on your behalf (MCP’s – more on that later)
• Estate and Letting Agency is in the dark ages when it comes to the experience the customer can enjoy (in particular in self-service)

In rare moments of peace I’ve been thinking about these two elements that are leading me to reimagine what

What is an MCP (Model Context Protocol)?

Most of us will have heard of an API, the way that two software programs can exchange information.   The MCP is built on API’s, but with a big difference in capability in the age of AI.

MCP vs API — What’s the Difference?

API (Application Programming Interface) is the foundational concept — it’s a defined contract that lets one piece of software talk to another. When you call a weather API, your app sends a request in a specific format and gets data back. APIs are general-purpose and can be used by anything that knows the spec.

MCP (Model Context Protocol) is a newer, more specific standard built on top of that same idea, but designed specifically for AI assistants like me. Anthropic introduced it to solve a particular problem: how do you give an AI model a consistent, safe way to use external tools and services?

Here’s a useful way to think about the difference:

A regular API is like a telephone switchboard — any caller who knows the number and speaks the right language can connect. It’s designed for software-to-software communication in general.

An MCP server is like a trained assistant who already knows how to use that switchboard on your behalf. It wraps the underlying API in a standardised way that an AI model can understand, with built-in descriptions of what each tool does, what parameters it needs, and what it returns — all in a format the AI can reason about.

How can I try and MCP today?

If you have started to use Claude (which I would highly recommend) then you have access to a huge range of MCPs today

• Gmail
• Google Drive
• Hubspot
• Slack
• Notion 
• And hundreds of others.

What it means is that using natural language in Claude you can turn your AI chat into action.  You can get Claude to research a topic and then send an email.  

You can have Claude read all your e-mails on a specific topic, read notes from a meeting in Notion and get files from Google Drive and consider all of those topics to build a spreadsheet from scratch and place it in your google drive.

MCPs in the Context of Estate and Letting Agency Software

Consumer Facing

Greenhouse OS will be the first Estate and Letting Agency software to have a fully independent MCP server.  We will allow customers to be able to book viewings and valuations by interacting with your database with natural language conversation. 

This will transform lead handling from a forms based product or something which sits as an ‘interceptor’ between the lead and your CRM.   Instead it will become a qualifying registration system that only allows customers to book viewings if they meet specific criteria.

In the current versions of Greenhouse OS, this journey is linear.  You start off filling in forms and you end booking a viewing.  Not too shabby when you consider 38% of our viewings are booked by the customer.

But what is there is a golden MA opportunity, the current version can’t pitch, close and book it.  That’s done by a different AI agent.  This system changes everything.  The AI can now intervene when it becomes appropriate.  That works in both directions.  It contacts me to see if I want to switch agents and I am interested in booking a viewing, it can now handle that and access those API’s as well.

As I write the minimum wage is about to go up again, and certainly prices are rising again. Every business is going to be affected.   We have to find ways to deliver a better service at a lower cost.

This gives a huge step in that direction, for the businesses that are able to access and capitialise on this technology.

Agent Facing

What MCPs were used to allow agents to talk to their CRM to get actions executed?

You are out on the road and you want to arrange a second viewing or record an offer and you just tell the software what you’ve agreed.

This next phase of MCPs will be the new battle ground in ‘user experience’ as we blend easy ways for customers to add themselves, at the same time as making it easier for agents to execute actions so that all the data is recorded in the ‘single source of truth’.  

This allows agentic AI to be vastly more powerful.   With all of the information recorded from phone conversations directly into the CRM, where we choose to, the Agent’s we customise and create will have access to all of the information that has occurred in telephone conversations, emails and whatsapps in one place.  Its timing and content of communications now have the capacity to be perfect.

The agents who master this fusion between AI agents, making it easy for their customers and easy for their teams

I can be fairly certain it won’t work out exactly like this.  Things are changing too fast.   By the time we have executed this, new, as yet unimagined possibilities will have emerged.  What is certain is that a static CRM that powers your world without the ability for you to change fast is going to become even more dangerous than it is today.

If you agree or disagree with this article – feel free to let me know.

I will put my Whatsapp below, so you can drop me a message if you want to have a conversation about how far we are on this journey with Greenhouse OS