Connect Claude to Your Freelance Business: The MCP Setup Guide

What MCP actually is, in one paragraph

MCP (Model Context Protocol) is a way for your AI assistant to actually see and act on your real data instead of you copy-pasting into a chat window. Without it, asking Claude "what's overdue from last month" means you first have to export a list, paste it in, and hope you didn't leave anything out. With MCP, Claude can ask your invoicing system directly and get a current answer — because it's connected to the real thing, not a snapshot you fed it.

That's the whole idea. Everything below is just the mechanics of turning that connection on.

Before you start

You'll need a Zenaris account and an MCP-capable client — Claude Desktop, Claude Code, or any other tool that supports connecting to remote MCP servers (ChatGPT has been rolling out support too). You don't need an API key, and you don't need to be a developer. The whole setup happens through a normal browser sign-in.

Step 1: Add the Zenaris MCP server

In your MCP client's settings, look for where it lets you add a remote server or connector, and add:

https://www.zenaris.co/api/mcp

Exactly where this setting lives depends on which client you're using and tends to move around as these tools update, so if the menu looks different from what you expected, search your client's settings for "MCP," "connectors," or "remote servers."

Step 2: Sign in through the browser

Your MCP client will open a browser window and walk you through Zenaris's normal sign-in flow — the same one-time-code login you'd use on the website. There's no separate password or API token to generate and paste around. Once you approve the connection, the browser tab closes and your client is authenticated.

Step 3: Try it

That's it — setup is done. A few prompts that actually work once you're connected:

• "Log 3 hours to the Acme project."
• "What's still unpaid from Q2?"
• "Summarize my notes on the Acme kickoff call."
• "What tasks are still open for Meridian Consulting?"

These aren't hypothetical demo prompts — they're the kind of thing that's genuinely faster to say out loud than to go find in an app, especially the ones you'd normally do right after a client call while the details are still fresh.

You can also chain these together in one conversation. "Log 2 hours to the Meridian redesign, then tell me what's still outstanding on their account" runs as two tool calls in sequence, and you get back a single answer instead of switching between a timer and a dashboard to check both things yourself. That's a small example, but it's the shape of most of the genuinely useful prompts — less "ask a single question," more "handle this whole little errand for me."

What's read-only vs. what needs your confirmation

This is the question worth asking before you connect anything to your business data, so here's the direct answer: reading is unrestricted, writing is not automatic.

Claude (or any connected assistant) can read your invoices, clients, tasks, notes, and time entries freely — that's how it answers questions like "what's overdue." Actions that change something, like logging a time entry or creating a task, go through your MCP client's own confirmation step before they happen. Zenaris never auto-sends anything on your behalf through MCP, the same way it never auto-sends client emails through the rest of the product. If an action would send something to a client or spend money, it requires your explicit approval in the moment, not a blanket permission you granted once during setup.

This isn't a limitation bolted on for MCP specifically — it's the same propose-not-send principle Zenaris uses everywhere else. The AI can draft, suggest, and surface information. It doesn't act behind your back.

A few questions worth asking before you connect anything

What if I use ChatGPT instead of Claude? The setup is the same — add the server URL in ChatGPT's connector settings and sign in the same way. MCP is a protocol, not a Claude-specific feature, so the steps above should generalize to any client that supports remote MCP servers, even ones that don't exist yet.

What can it actually see? Whatever your Zenaris account can see — your own clients, invoices, tasks, notes, and time entries. It's scoped to your account the same way logging into the website is; it can't see other users' data, and it doesn't get broader access than you already have.

What happens if I disconnect it? Nothing retroactive. Revoking access in your MCP client (or from Zenaris settings) immediately stops the connection. Nothing the assistant read or logged while connected gets undone, but no further access is possible until you reconnect.

Does this cost extra? No — MCP access is included, not a separate add-on tier.

Where this goes next

Right now, a connected assistant can log time, create tasks, and search across your invoices, clients, and notes. As the tool surface grows, expect more of your day-to-day admin — the stuff you'd normally do in five clicks across two apps — to collapse into a single sentence you say while you're already thinking about the work.

Get started

If you haven't set up a Zenaris account yet, you can do that first and connect MCP right after — see the AI assistant / MCP feature page for more detail, or check pricing if you're deciding which plan to start on. For the bigger-picture argument on why this is worth doing now, see The Future of Freelancing Is Agentic.