MCP (Model Context Protocol)

neural network diagram

What is MCP?

The Model Context Protocol (MCP) is an open standard for connecting AI assistants to the systems, data and tools they need to do useful work. It was introduced by Anthropic in late 2024, and is now supported by the major AI assistants and developer tools.

Before MCP, every AI product needed its own custom integration with every system it talked to. MCP replaces that with one common connection: a system that offers an MCP server can be used by any assistant that supports MCP — much as a USB port lets any device plug into any computer.

How it works

An MCP server sits beside a business system and describes what an AI may do with it:

  • Tools — actions the AI can take, such as looking up a customer, creating a task or checking stock.
  • Resources — information the AI can read, such as records, documents or reports.
  • Prompts — ready-made instructions for common jobs, so results are consistent.

An AI assistant, such as Claude or ChatGPT, connects to the server, discovers the tools and data on offer, and uses them automatically whenever a request calls for it. You stay in control: your system decides exactly what is exposed, and to whom.

Build one MCP server for your system, and any MCP-compatible AI assistant can put it to work, today and as new ones arrive.

MCP & your business

MCP lets your staff use the AI assistants they already have with your own systems, rather than copying information back and forth. For example, they could:

  • ask which jobs are running over budget this month, and get an answer drawn from your live data
  • have an assistant summarise a customer’s history before a meeting
  • draft quotes, tasks or replies straight into the right system, ready for a person to check

Because MCP is a shared standard, the same server keeps working as AI assistants improve or as you change between them.

Doing it safely

Giving AI access to business systems needs the same care as giving access to a new staff member. We design MCP servers so that:

  • the AI inherits each user’s permissions, so it can only see and do what that person is allowed to
  • access begins read-only, with the ability to make changes added step by step, and a person confirming any action that matters
  • every request is logged, giving you a complete audit trail of what was asked and what was done
  • sensitive data can be excluded or masked, so it is never sent to the AI service

What does it mean for us?

At Canary, we design and build secure MCP servers for both new and long-established business systems, typically in C# and .NET using the official SDK, so they fit in with the systems we already develop and support. See our AI Powered page for how this fits into our wider approach.

Get started

Tell us which systems your team would most like their AI assistants to work with. We’ll tell you what an MCP server could safely offer, and what it would take to build.

Talk to us about MCP

Read more about how we build software in our software development section.

How can we help?

Use this form, or email info@canary.co.nz

Name

Email

Company (if applicable)

Phone

Enquiry