Skip to content
// Glossary · Protocol and platform

Model Context Protocol (MCP)

Also called: MCP

What is the Model Context Protocol?

The Model Context Protocol is an open standard that lets an AI agent discover and call external tools in a uniform way. Instead of writing a separate integration for every client, a tool server describes what it offers, and any MCP-compatible agent can use those tools directly.

// In more depth

What the protocol actually does.

MCP solves a combinatorial problem. Without it, every tool needs an integration for every client, and the number of integrations grows by multiplication.

01

A server describes its tools

An MCP server publishes the tools it offers, what each one does, and what it expects. That description is machine-readable, so the agent does not need a human explaining the interface.

02

The agent decides when to call them

Rather than a person invoking a feature, the agent recognises that a tool is relevant mid-task and calls it. This is the difference between a tool you use and a tool your agent uses.

03

Any compatible client works

Because the convention is shared, one server serves every MCP-compatible client. A tool built today works with a client that ships next year, without either side coordinating.

// Why it matters

Why it matters for agent tooling.

The protocol is what makes a capability portable across the tools a team actually runs.

Portability across clients

Teams rarely standardise on one coding agent. A capability behind MCP reaches all of them rather than whichever one it was written for.

Durability over time

Tool preferences change faster than the knowledge and workflows built around them. A protocol outlives the client of the moment.

Agent-initiated, not human-initiated

The agent calls the tool when it needs it, so capability is available mid-task rather than requiring somebody to stop and paste context.

// Commonly confused with

What MCP is not.

Three distinctions that come up constantly.

Not a modelMCP does not run or provide an AI model. It is the calling convention between an agent and the tools it uses.
Not a security boundaryThe protocol says nothing about who can access what. Access control is the responsibility of the server. Virex enforces it by organization and by seat.
Not vendor-specificMCP is an open standard with multiple independent implementations, rather than a proprietary interface for one product.
// FAQ

MCP questions

Who created the Model Context Protocol?

MCP is an open protocol published by Anthropic and adopted across a range of AI coding tools and agent frameworks. It is not a Virex invention, which is precisely why building on it is useful.

What is an MCP server?

A service that exposes a set of tools with descriptions an agent can read. The agent decides when to call them. Virex ships one for Virex Memory and one for Virex Remote.

Does MCP mean my data is shared with the tool vendor?

No, and it says nothing about that either way. MCP is a calling convention. What a given server does with your data depends on that server. With Virex, calls resolve inside your own organization only.

Do I need MCP to use Virex?

No. Both products have full web interfaces. MCP is how your agents use them without a person in the loop.

Is MCP the same as an API?

Related but not identical. An API is an interface a developer writes code against. MCP is a convention for describing tools so an agent can pick them up and decide to use them on its own.

Give your agents the tools.

Subscribe, assign a seat, and the MCP surface for that product is available to whoever holds it.