What is MCP for business software?

The Model Context Protocol (MCP) is an open standard that lets AI assistants read from and act on business software through a defined set of tools, instead of copied data or one-off integrations.

Quick answer

MCP, the Model Context Protocol, is an open standard for connecting AI clients to external tools and data. A business system exposes a server with named tools, such as listing services or preparing a draft, and the AI client calls those tools within the permissions the user has granted.

Who builds the connection

An MCP server sits in front of a business system and describes what it can do. An MCP client, such as an AI assistant or a coding tool, connects to that server and decides when to call one of its tools.

What a tool is

Each tool has a name, the inputs it accepts and a result it returns. Read-only tools look information up. Tools that create, change or remove records need more care, because the result is a real change in the business system.

How access is granted

A business server should sign the user in and limit each connection to the data that user may see. Komashi uses OAuth sign-in with a scope for MCP use, and it checks each call against the signed-in account.

Where it helps and where it does not

MCP removes copy-and-paste between tools when a task needs live data. It does not make business decisions for you. An AI answer is not an approval for a price, a customer or a cancellation.

When MCP is useful

  • Looking up current services, plans or customers while you work in an AI tool.
  • Preparing a draft that a person reviews before anything goes live.
  • Keeping one set of permissions for every AI client that connects to the system.

A practical example

An owner of a small maintenance business asks their AI assistant which plans have no yearly option. The assistant calls a read-only tool that lists services and packages, then names the three plans without a yearly price. The owner then asks for a draft yearly package. The assistant returns a summary to review, and nothing is created until the owner approves it in Komashi. This example is illustrative, not a customer case.

Official source

The protocol definition and client guidance are published by the Model Context Protocol project.

How Komashi fits

Komashi runs an MCP server for subscription, service, customer and usage data. Connections sign in with OAuth and the mcp:use scope, every call is checked against the signed-in account, and setup requests return drafts for you to review. The developer page lists the connection steps and the full tool set.

Read the MCP developer guide

What to check before you connect an AI tool

  • Which data the connection can read, and whether that matches what your team expects it to see.
  • Which tools change or remove data, and who approves those changes.
  • How to revoke a connection when a tool or device is no longer in use.
  • That the AI tool supports remote MCP servers before you plan around it.

Ready to put this into practice?

Create a service with Komashi.