Your Terminal Is The New Marketing Department

TL;DR: People are now using Claude Code to build full marketing systems from the terminal. Research, copywriting, landing pages, ad variations, automated reporting. This article breaks down the frameworks, tools, and workflows that are making it possible, based on the work of marketing experts.

For most founders, the dream is shipping code. The nightmare is everything after that.

You build a great product. Then you spend months trying to find your first customer. Maybe you hire an agency that takes weeks to deliver a landing page and a few ads.

The gap between building and selling has never been wider, but that gap is closing fast.

We’re entering what James Dickerson calls an era of “Vibe Marketing.”

The Vision

What if you could build your marketing systems in the same place you build your product? What if research, copywriting, design, and deployment could all be supported by trained AI agents?

I’ve been studying how marketing experts are using Claude Code to better understand what’s useful vs. what’s hype. This article can help set the foundation for anyone who wants to learn more and start taking action.

The WAT Framework

WAT stands for Workflows, Agents, and Tools. It sounds technical, but the concept is pretty simple.

Workflows = Your standard operating procedures written as markdown files. Think of them as playbooks the AI follows. “Here’s how we write a case study.” “Here’s how we research a new market.” You write these once and the AI references them every time.

Agents = The reasoning layer. This is Claude Code, figuring out what to do next based on the workflow you gave it. It reads the situation, decides on the right approach, and coordinates the work.

Tools = Scripts that actually do things. Send an email. Scrape a website. Pull data from Google Sheets. The tools execute, the agent thinks, and the workflow keeps everything on track.

This is what separates useful AI marketing from the generic “write me a blog post” prompts that everyone’s been using.

MCPs: The Connectors That Make It All Work

The Model Context Protocol (MCP) is probably the most important concept here, and the least understood. Allie Miller describes it as a “universal port” for your business. I think that’s the right way to frame it.

MCPs let Claude Code plug into your existing tools. Your Gmail. Your Google Sheets. Your browser. Your research tools.

A few that are worth knowing:

• Perplexity MCP connects Claude to deep market research. Instead of guessing what your competitors are missing, you can identify gaps. One example from Dickerson’s work: he found that local plumbers were losing leads simply because they didn’t respond fast enough. That insight became the positioning for a campaign.
• Playwright MCP gives Claude the ability to automate browser actions. It can navigate websites, capture screenshots, and pull information without switching between tabs.
• Google Workspace MCP connects your Gmail, Sheets, and Docs. Anthropic built their own version of this, and if you’re connecting sensitive data, that matters. Security should be the first conversation, not an afterthought.

Skills: Teaching AI to Think Like Your Best People

This is where things get interesting.

Dickerson built a system of 17 “Skills” for Claude Code. Each one is basically an instruction manual for a specific marketing discipline. Direct response copywriting. Competitive positioning. Frontend design. Each skill is loaded with frameworks, references, and guardrails so the AI doesn’t default to generic output.

The one that ties everything together is the Orchestrator Skill. It’s a master agent that looks at your project, figures out where you are in the process, and decides which skill to run next. You’re not prompting every time. The system knows what comes next.

He also built a Frontend Design skill that deliberately avoids what he calls “AI slop.” No purple gradients. No rounded corners everywhere. No emoji-heavy layouts. The goal is a professional, conversion-focused design that doesn’t look like it was generated by a chatbot.

Real-Time Context with “Last 30 Days”

Matt Van Horn built a skill called “Last 30 Days” that solves one of AI’s biggest weaknesses: stale data.

It pulls trending content from X and Reddit and feeds it into your session. So when you’re writing hooks or developing campaign angles, you’re working with what’s actually resonating right now. Not what was trending when the model was last trained.

This is the difference between writing copy that feels current and writing copy that feels six months old.

From Creator to Scientist

Traditional agencies sell you one landing page. One version. One creative direction. You’re betting everything on taste and intuition.

With tools like Remotion (programmatic video editing) and the ability to generate variations at scale, you can build 100 landing pages targeting 100 different keywords. You can create dozens of video ad variations in a single session.

Your job stops being “make the thing” and starts being “test the thing.” Deploy variations, see what the market responds to, and double down on winners.

Allie Miller talks about “Command Stacking” as the next evolution. You set up a /dailybrief command that runs every morning. It checks your calendar, preps background on every meeting, pulls relevant threads from Slack and Gmail, and has everything waiting for you before your first coffee. You can set up automated lead magnets that capture information and populate a Google Sheet in the background while you sleep.

What This Means for Marketers

To be clear, all this automation doesn’t mean taste and strategy don’t matter. They matter more than ever.

Marketers who win have real opinions about positioning, messaging, and audience. The AI handles execution. You bring the perspective. The role this work points at is the one I wrote about in the rise of the marketing engineer: a generalist with taste, plus the tools to ship.

What’s changing is the cost and timeline of going from insight to live campaign. What used to take a team of 10 people and 8 weeks can now happen in days with one person who understands the tools. The fastest way to get there is the way I outlined in stop learning AI alone: build a small circle of operators and trade workflows weekly.

The question isn’t whether AI will change marketing operations. It already has. The question is whether you’re building the systems to take advantage of it, or waiting for someone else to figure it out first. If you want help mapping where AI fits in your business, the AI Opportunity Sprint is the four-week walk-through.