For business For enterprise Solutions Apps Pricing Developers Docs Launch a workspace
Resources / Guide

What is the Model Context Protocol?

For an AI agent to be useful inside a business, it needs to do more than talk. It needs to reach into the tools the business runs on and act. The Model Context Protocol is the open standard that makes that possible, the same way a common plug lets any appliance use any socket.

The one-sentence version

MCP is an open standard for connecting AI agents to tools and data. A piece of software exposes its capabilities as an MCP server; any MCP-capable client, the agent you already use, connects to it and can then call those capabilities directly.

Why a standard was needed

Before a common protocol, every connection between an agent and a tool was a bespoke integration: build it, maintain it, and rebuild it when either side changed. That does not scale. A shared standard turns a tangle of one-off integrations into a single way of connecting, so an agent that speaks MCP can work with any tool that speaks MCP, with no glue code in between.

How it works, briefly

There are two sides. A server publishes a set of tools, each one a concrete action with a clear description and inputs. A client, the agent, discovers those tools and decides which to call to achieve the goal you gave it. The protocol carries the request and the result. Because the tools are described in a structured way, the agent can reason about which to use rather than being hard-wired to a script.

What it means for you

Two things. First, you are not locked to one vendor's assistant: any compatible agent can connect, so you bring the one you trust and swap in a better one whenever the field moves. Second, the agent can do real work, because everything a person can do is published as a tool it can call, not a narrow set someone decided to wire up.

MCP and SOIS

A SOIS workspace is a full MCP server. Every action a person can take, across contacts, finance, the inbox, stock, and the rest, is available to your agent over the protocol, within your permissions and your budget. That is what makes it agent-native rather than a screens-first product with AI added on. See how to connect your agent, or read the glossary for the rest of the vocabulary.