Work with your AI agent

What is MCP? A plain explanation for agency owners

MCP lets your AI agent use your other programs and files directly. What it is, what it changes for a one-person AI agency, and what to check first.

By Volant team

If you use an AI agent for client work, you have probably copied things into a chat: a brief, a spreadsheet of businesses to pitch, notes from a call. MCP is the reason you more and more don't have to.

This article explains what MCP is, what it changes for a one-person AI agency, and what to check before you connect anything to client information.

What is MCP?

MCP, short for Model Context Protocol, is a shared set of rules for connecting AI programs to other programs. The official documentation describes it as a way for AI programs to reach information, tools and everyday tasks, and compares it to the USB-C plug on a phone or laptop: one kind of plug instead of a different cable for every device.

In practice, a small program called an MCP server offers something, such as your files, a calendar or a list of customers, and any AI program that supports MCP can connect to it and use it.

Who made MCP, and who supports it?

Anthropic, the company behind Claude, introduced MCP and made it free for anyone to use on 25 November 2024. Its announcement named the problem it set out to solve: every new source of information needed its own custom connection.

Since then, several AI assistants and developer tools have added support. The MCP documentation lists Claude, ChatGPT, Visual Studio Code and Cursor among the programs that support it. Check the current documentation before you rely on a specific one, because support changes.

What does MCP change for agency work?

It changes where your agent gets its information. Without a connection, your agent knows only what you paste into the chat, and it forgets it when the chat ends. With a connection, it can read the notes you keep, such as a client brief or a list of businesses to pitch, and work from those.

Claude Code's documentation puts it plainly: add a connection when you find yourself copying information into the chat from another program.

For a one-person agency, that means:

  • your client notes can live in one place, outside any single chat;
  • the agent can read the latest version of a brief instead of an old pasted copy;
  • the work the agent does can be saved back in the same place.

What should I check before connecting client information?

MCP makes it easier for an agent to read and act on your programs. That is the point, and it is also the risk.

  • Trust each connection. Claude Code's documentation tells you to make sure you trust each one before connecting it, and warns that a program that pulls in outside content can pass on hidden instructions meant to trick your agent.
  • Give the least access that works. A connection that only needs to read a folder should not be able to change it.
  • Stay in charge. The MCP security principles say people must clearly agree to, and understand, what information is used and what is done with it, and must stay in control of what is shared and what actions are taken.
  • Keep clients apart. Don't give an agent working on one client's site access to another client's files.

Does MCP mean my agent can send things on its own?

No, not by itself. It depends on the programs you connect and the permissions you give. MCP is a way to connect; it does not decide what should happen. A connected email program could send email. Whether it waits for your approval is decided by that program and your settings, not by MCP.

For anything a client or customer will see, such as sending an email, publishing a page or charging a card, keep a step where a person approves it, and make the approval cover one exact version of what goes out. That is a good working rule whatever tools you use.

How does Volant use MCP?

MCP is the connection that lets your agent use Volant, and how approvals work in Volant builds on the rule above, so your agent can read and update your client details and notes without you pasting them into a chat. The agent it connects is Claude Code.

It is worth saying plainly where Volant's part ends. Volant checks everything it sends, publishes or charges. Tools your agent uses on its own are outside Volant.

Questions

Questions people ask

What is MCP in simple terms?
MCP, short for Model Context Protocol, is a shared set of rules that lets an AI program connect to your other programs and information, such as files, a calendar or a list of customers. It is one common kind of connection instead of a custom one for each.
Who created MCP?
Anthropic, the company behind Claude, introduced MCP and made it free for anyone to use on 25 November 2024. Several AI programs and developer tools now support it.
Is it safe to connect my AI agent to client information through MCP?
Only as safe as the programs you connect and the permissions you give. Claude Code's documentation says to make sure you trust each one before connecting it, because a program that pulls in outside content can pass on hidden instructions meant to trick your agent.
Do I need to be a developer to use MCP?
No, not always. Many AI programs let you add a connection in a short setup step. Knowing what each connection can see and do is still your job.

Start here

Try one ideaon your next client.

Every how-to here works with a simple notes file. Keep the ones that help, and let Volant keep track when you're ready.

For Mac and Windows. Works with Claude Code.

Join the waitlist