I'm sorry. That was a cheap hook. I have no idea how much more productive you'll be, but what I can tell you is that MCP is a productivity game-changer. What it stands for (Model Context Protocol if you must know) is less important than what it actually does. And to figure out the latter, I spoke to Reggie Tan, a Myosin OG whom you might remember from our article on how Myosin members are using AI. 

MCP lets AI models like Claude directly use your tools: APIs, databases, browsers, and more. Instead of asking the AI to explain how to do something, MCP lets it actually do it.

In this post, we'll cover two ways MCP can transform your workflow as a crypto marketer. First, how to draft emails in your own writing style. Second, how to replicate your Twitter voice. These are the use cases Reggie and I have tested, and they work.

Projects: Your Marketing Command Centers

First, you'll need a Claude Pro account, which will set you back around $20-25/month. That's the bad news.

Once you're signed up, go to the left hand column and you'll see the option to create a Project. Without Projects, everything you do in Claude exists in one giant conversation blob. Claude has no memory between sessions, so every time you start a new chat, you'd have to re-explain everything. Projects fix that.

Think of them as command centers. Each one acts as a silo where Claude can reference specific files and instructions. You can create one for each client or function:

• "Ghostwriter: Client X"

• "Ad Performance Reports"

• "Content: Crypto Twitter"

When you're in a project folder, you'll see a section marked 'Files' on the right hand side. This is where you upload all the data relevant to that project: PDFs, Google Drive docs, Github code, whatever you need. Or you can just add files as text, which is how Reggie inputted his writing style.

Inside each Project, you want two things:

Instructions: Tell Claude what role it's playing. Example: "You're an expert ghostwriter working directly for me. No fluff, no excessive emojis, no AI-sounding phrases like 'landscape' or 'delve into.'"

Files: Upload style guides, strategy docs, and reference material. Claude will automatically reference these when responding.

Pro tip from Reggie: Use Claude's Deep Research mode when setting up or analyzing Projects. It takes longer, but the results are way more detailed. You get 20 deep research queries a month, so use them on the important stuff.

Copy Your Email Writing Style

Here's where the MCP magic comes in. MCP allows Claude to interact directly with any tools you connect to it. You do this by selecting 'settings' (left hand column) and then 'connectors'.

The MCP ecosystem connects AI tools to more than 500 business and productivity apps, and the possibilities are pretty wild. We'll get to that. But for now, make sure Gmail is connected.

Once Gmail is connected, go back to your project (left column) and ask Claude:

"Analyze my Gmail and describe my email writing style in detail."

Claude will come back with a very lengthy description of your writing style. It's pretty revealing. It will tell you how long, short, verbose, or succinct your typical email responses are. It will pick up on the nuance of your language, your classic sign-offs and turns of phrase. It can be eye-opening, jarring even, but it is what it is.

What you do next is important. Copy Claude's description, then in the same project folder (which you've titled 'Ghostwriter' or something more creative) select the add file option (right hand column), give it a title like 'Email Ghostwriter' and paste the writing style description as plain text. This document serves as the blueprint for your tone, so edit it until it reflects how you actually want to sound.

Once that's saved, Claude will automatically reference this file when you're working in this Project. Now you can prompt it with something like:

"Look through my Gmail, find the latest email from Satoshi and write a draft reply in my voice."

Claude will draft it for you, pulling from your Gmail and matching your style.

Important note: The LLM will get you about 90% of the way there. You'll still need to review the output, especially early on. When you spot things that don't sound quite right, add those corrections back into your writing style file. The more you refine it, the better Claude gets at mimicking your voice.

Copy Your X/Twitter Writing Style

This one uses a slightly different approach. We're switching to Grok here because it's simply better at analyzing X posts than Claude. And when you're training an AI on your voice, you want the best analysis possible.

Go to your X profile, open Grok, and give it this prompt:

Grok will crawl your entire posting history and generate a detailed style guide based on how you actually tweet.

Pro tip from Reggie: Use Deep Research mode in Grok for this. It takes longer, but the output is significantly more detailed. You get 20 Deep Research queries per month, so save them for tasks like this where depth matters.

Also, ask Grok to output the style guide in Markdown format. LLMs process Markdown far better than plain text because it preserves formatting, headers, and tone markers.

Once you have the guide, create a new Project in Claude (something like 'Crypto Twitter') and upload your Markdown style guide to the Files section. We're back to Claude now because it's the best writing model out there. Also, you don't want to sound like everyone using ChatGPT.

Now you can ask Claude:

"Write three posts in my X style about [topic]."

And it'll sound like you. Sometimes uncomfortably so.

Tip of the Iceberg

I showed you how to replicate your voice on Gmail and X. But MCP is far more powerful than that. Those 500+ app integrations I mentioned earlier? Here's what becomes possible when you start combining tools.

You could connect Claude to your analytics platforms, project management tools, and databases. Have it monitor Discord channels and Telegram groups for your daily brief. Wire it up to pull Twitter analytics, cross-reference your CRM data, and draft performance summaries in your voice. Or have it watch your Postgres database, detect when metrics hit thresholds, generate reports, and post them to Slack. No code required.

Email and Twitter are just the warm-up. Once you see how to connect tools and feed Claude your context, you’ll start spotting automations everywhere. It takes just an afternoon to set up and will quietly save you hours every week after that. The power of AI is compounding.