Skip to content
← Back to Journal
Trends
6 min read

MCP Explained for Your Business (No Coding Required)

MCP is the USB-C port between your AI and your business software. Here's what it is, what it already does today, and why your ERP needs to speak it.

By Frihet Team

TL;DR: MCP is an open standard that lets an AI connect to your business tools in a structured way, instead of staying locked out. Frihet already runs a real MCP server: today you can ask an agent like Claude to pull up your invoices, create a new one, or look up a client. An ERP without MCP sits outside this shift.

Share
MCP Explained for Your Business (No Coding Required)

Key takeaways

  • MCP solves the N×M integration problem: without it, every AI needs a custom build for every tool it touches.
  • Frihet exposes a live MCP server at mcp.frihet.io — an agent can read invoices, create one, and search clients today, not on a roadmap slide.
  • There is a real gap between "AI connected to your data" (live today, human still confirms) and "AI running your business unsupervised" (a direction, not a current feature).
  • An ERP without a public API or MCP support cannot participate in the agentic economy, no matter how polished its UI is.
  • You don't need to write code to benefit — you need your business software to speak the protocol.
Contents

The problem isn’t the AI. It’s that it can’t touch your business

You have an AI assistant that writes better than most of your team, reasons through numbers, and drafts in seconds what takes you an hour.

And yet, to invoice a client, you still open a tab, fill out a form, and click save.

The AI is brilliant and locked out. It can reason but it can’t touch anything — not your ERP, not your bank, not your calendar. Every time you want it to act on real data, you have to be the bridge yourself. Copy, paste, confirm.

That’s the problem MCP solves. It doesn’t make the AI smarter. It gives it hands.

MCP in three sentences

Model Context Protocol (MCP) is an open standard that defines how an AI connects to external tools and data in a structured way.

Instead of every application building its own custom integration for every AI assistant that exists, you build a single MCP server once, and any compatible AI knows how to talk to it.

It turns the N×M integration problem into N+M: without a standard, you need one integration per tool-per-assistant combination; with one, you need a single integration per tool, period.

What changes when your ERP speaks MCP

Without MCP, your ERP is an island. It holds all your financial information — invoices, expenses, clients, margins — but only you can go in and dig it out, field by field.

With MCP, that same information and those same actions become available to an agent you authorize. You ask for something in a sentence, and the agent executes the action through the same system, with the same permissions you’d have yourself.

A concrete example: you tell Claude, connected to your Frihet account, “invoice Mendoza Studio €620 for the June branding project.” The agent identifies the client, builds the invoice with your correct tax details, and creates it. You review it and send it. What used to be four screens and two minutes is now one sentence.

Or the reverse, querying instead of executing: “which invoices are more than 30 days overdue?” The agent checks your invoices in real time and answers with the list, without you opening anything.

What Frihet’s MCP does today, and what’s direction

This is where precision matters, because there’s a lot of noise in this space.

Frihet is a Spanish ERP with an MCP server, a documented public API and a free tier. Its MCP server, at mcp.frihet.io, isn’t a product promise or an internal demo. It’s a live feature you can use today by connecting a compatible agent like Claude.

What works right now:

What is not a live feature today: an agent that invoices, collects payment, and reconciles your books without you stepping in at any point. That’s the direction the so-called “agentic economy” is heading — businesses where part of the operation runs on autopilot, supervised but not hand-executed by a human. It’s the horizon, not this week’s product.

Why an ERP without MCP is sitting out this shift

Here’s the bet, stated plainly: AI-agent adoption in day-to-day work keeps growing, not the other way around.

Every month, more people delegate routine tasks — drafting, summarizing, searching, comparing — to an assistant. It’s only a matter of time before that delegation extends to running the business too: checking numbers, generating documents, executing repetitive actions.

When that happens, software that already has the door open saves time. Software that doesn’t forces its users to keep copying and pasting manually between their AI and their ERP — or to skip the AI for that part of the work entirely because there’s no safe way to connect it.

The honest counter-argument: maybe you don’t use any AI agent to run your business today. That’s a reasonable position — not everyone needs this yet, and pretending otherwise would be selling hype. But the point isn’t “use it now”; it’s that your business software should be able to do this the moment you decide you need it — the same way not everyone needed an API in 2010, but software without one got left behind once integrations started to matter.

A closed ERP, without a public API or MCP, isn’t a neutral choice. It’s a bet that this shift never arrives. We’re not making that bet.

The point that matters

MCP isn’t a niche feature for developers. It’s the difference between an ERP your AI can actually use and one your AI can only describe secondhand.

Frihet bet on this before it was obvious: a real MCP server, a documented public API, and a free tier so nobody has to risk money to try it. Running your business — whether you invoice ten euros or ten million — should be ready for an agent to operate it with you, not despite you.

The rest of the market will catch up. In the meantime, the door is already open.

Was this article helpful?

FAQ

Do I need to know how to code to use MCP with Frihet?

No. MCP is the plumbing underneath — you just talk to your agent in plain language ("invoice this client for €400"). The technical part — the agent knowing which tools it has and how to call them — is already solved by Frihet's MCP server.

What can an agent connected to Frihet MCP actually do today?

Read your invoices, create a new invoice, look up a client, or query your business data through an agent like Claude. This is a live feature, not a roadmap announcement. What it does not do yet is run your business unsupervised — that's a direction, not a shipped product.

Does MCP replace Frihet's public API?

No, it complements it. The API is for developers building custom integrations. MCP is for an AI to use that same capability conversationally, without anyone writing integration code for each individual assistant.

Compare with

Related articles

Comments

Frihet — Business without drama

Start Free