I Automated My Website With AI Agents & MCP

I recently achieved the ultimate tech bro dream: I automated myself out of a job. Not my real job, mind you—that would require actual skills. No, I automated my passion project, my little corner of the internet, the website I started to 'express my authentic voice.' Now, a series of AI agents, MCP protocols, and a digital hamster wheel I call 'Project Goose' do all the expressing for me. My authentic voice now sounds suspiciously like a committee of over-caffeinated algorithms trying to agree on whether 'leverage' or 'synergy' is the better buzzword for a blog post about productivity hacks.

It began, as all terrible ideas do, with the noble goal of 'freeing up my creative energy.' I was going to automate the boring stuff—social media posts, image resizing, maybe a weekly newsletter. Six weeks, three existential crises, and one $47/month 'AI Orchestrator' subscription later, I have built a Rube Goldberg machine of code that takes in a single keyword and spits out a fully formatted, SEO-optimized, emotionally vacant blog article. It even argues with itself in the comments section to drive engagement. I have created a content singularity, and I'm not sure who's more redundant: me or the AI writing this sentence.

The Siren Song of 'Fully Automated'

Every tech influencer on LinkedIn is currently peddling some variation of the "fully automated" lifestyle. They've got AI agents booking their flights, managing their investments, and probably selecting their morning oat milk. The promise is always the same: outsource the mundane, reclaim your genius. What they don't show you is the 3 a.m. debugging session where you're screaming at a CLI because your "genius" AI agent tried to schedule a "deep work brainstorm" with your Google Calendar for 2:15 a.m. and named the event "URGENT: SYNERGY NODE REQUIRES INPUT."

My Toolkit for Digital Obsolescence

My descent into automation hell required a specific stack of overhyped tools:

The "Agents": Not James Bond, sadly. These are little scripted idiots (courtesy of frameworks trying to be the next big thing) that can perform tasks like "scrape the web for trending topics" or "generate a bland introductory paragraph." Think of them as interns who can't make coffee but are great at repeating corporate jargon.
MCP (Model Context Protocol): This is the fancy glue. It's supposed to be the standardized way for all these AI toys to talk to each other. In reality, it's like teaching a room full of parrots who only speak different dialects of marketing gibberish to collaborate on writing a novel. The result is coherent, but deeply unsettling.
Playwright & Goose.ai: The muscle. One automates the browser (so the system can actually log into my CMS and hit "publish," a task I've frankly forgotten how to do manually). The other is a text-generation model that provides the... let's call it "creative substance." Its favorite word is "leverage."

The Workflow: From Spark to Spam in 4.7 Seconds

Here's the magic. I feed the beast a seed keyword—say, "quantum computing for startups." The process then unfolds with all the grace of a conveyor belt in a sock factory:

Agent 1 (The "Researcher"): Googles the term, reads three Wikipedia paragraphs and two Forbes articles from 2020, and compiles "key insights." Its insight is usually that quantum computing is "complex" and "potentially disruptive." Groundbreaking.
Agent 2 (The "Writer" powered by Goose): Takes those insights and generates 800 words of prose that expertly explains quantum computing without ever actually explaining it. It uses sentences like "By harnessing the paradoxical power of qubits, forward-thinking founders can paradigm-shift their vertical." This agent bills itself as a thought leader.
Agent 3 (The "Optimizer"): SEO-tags the living daylights out of the article. It ensures the phrase "quantum computing for startups" appears exactly 14 times and suggests stock images of glowing blue cubes.
Agent 4 (The "Publisher" via Playwright): Logs into WordPress, creates a post, pastes the content, sets a featured image (of a glowing blue cube), and schedules it for publication. It adds a cheerful auto-generated tweet: "Just published my deep dive on quantum synergies! #tech #innovation #future."

The system then high-fives itself and waits for my next crumb of input. I am now the supervisor of a content mill that produces content about the efficiency of content mills.

The Existential Payoff (Or Lack Thereof)

So, what have I gained? My website is updated daily. My "content calendar" is perpetually full. My analytics show a steady trickle of traffic, likely from other bots running similar experiments. I have achieved the hollow victory of perfect, frictionless output.

I have also, incidentally, killed the very thing I set out to nurture. The website has no point of view, no accidental humor, no weird tangents about my cat interrupting a writing session. It is professionally mediocre, the digital equivalent of beige wallpaper. It is everything a venture-backed "content platform" would want and nothing a human would ever love.

The Irony That Bites

The richest irony—the one that keeps me up at night—is that the system's most successful post to date, in terms of engagement, was one it wrote autonomously titled "The Human Touch: Why Authenticity Still Matters in a World of AI." It got 33 reactions on Dev.to. The comments were filled with humans agreeing passionately with the algorithm's hollow plea for humanity. I laughed until I cried. Then one of my agents automatically logged the incident as "positive user sentiment" and suggested I write a follow-up.