The Hex Constraint: What It Is and Why It Exists
The hex is a constraint framework for building AI workflows. Six skills, six commands, one config. It maps directly to how Anthropic's Model Context Protocol architecture actually works — not as a metaphor, but as a design spec. The hex exists because the problem was never "which AI tools should I use." The problem was always "how do I stop collecting tools and start using them."
The Shape of the Thing
A hex is a hexagon. Six sides, six skills, one center. Each skill is a specific capability — drafting, image generation, email, scheduling, file management, publishing — and each one maps to a real tool connected through MCP. The commands are how you trigger those skills: verbs like /publish, /draft, /schedule. The config is a single file that holds your API keys and connection strings. That's the entire architecture. If you need a second page to describe your setup, you've already left hex territory.
The number six is not arbitrary, though it's also not sacred. Miller's Law — the cognitive psychology finding that working memory holds roughly seven plus or minus two chunks — provides the research basis. Six keeps you one slot below the median ceiling, which means you have cognitive room left over for the actual work the workflow is supposed to support. But the real reason is simpler than the research: six forces you to cut. You can't have six skills if three of them are aspirational. Every slot costs something, and that cost is what makes the hex useful.
When someone tells me their AI publishing workflow needs fourteen skills, I tell them they have two or three workflows pretending to be one. Separate them. Each one gets a hex. Suddenly you can think about each one clearly, test each one independently, and fix each one without breaking the others.
The Problem It Solves
There's a specific person this framework is built for — and if you're reading this, there's a decent chance it's you. You have a Claude subscription. You have a ChatGPT subscription. You probably have Notion, maybe Zapier or Make, possibly an n8n instance you set up one weekend and haven't touched since. You've bookmarked fifteen MCP servers on GitHub. You've watched hours of YouTube tutorials on AI automation. And the number of workflows you have running — actually running, on a schedule, producing output you use — is somewhere between zero and two.
This is the fourteen-subscription, zero-running-workflow trap. It's not a knowledge problem. You understand how these tools work. It's not a motivation problem. You genuinely want to build something. It's a constraint problem. You have too many options, too many tools, and no forcing function that says "pick six and go." The hex is that forcing function. It doesn't tell you which tools to use. It tells you how many, and that limit does most of the work.
The constraint also solves a subtler problem: the perpetual optimization loop. Without a fixed number of slots, every new tool launch becomes a question. "Should I add this?" "Should I swap that?" "Should I rebuild my whole setup around this new thing?" With a hex, the question becomes binary. "Is this better than what's currently in slot three?" If yes, swap. If no, move on. The decision surface shrinks from infinite to manageable.
How It Maps to MCP
The hex isn't just a productivity metaphor bolted onto a PDF. It maps directly to Anthropic's Model Context Protocol — the architecture that lets Claude connect to external tools, services, and data sources. Every skill in your hex corresponds to an MCP connector. Your email skill runs through an MCP server that gives Claude access to Gmail. Your file management skill runs through an MCP server that gives Claude access to your local filesystem. Your calendar skill, your publishing tool, your database — each one is a connector, and each connector is a slot in your hex.
This matters because MCP is the first time the constraint-based approach has had real infrastructure underneath it. Before MCP, "limit your tools" was willpower advice. You could decide to only use six tools, but they wouldn't talk to each other. You'd still be copying and pasting between tabs, manually bridging the gaps between services that couldn't see each other. MCP changes that. When Claude has access to your email, your calendar, and your files through MCP connectors, it can draft an email based on today's calendar and attach a file from your drive — in one conversation turn, without you touching a browser tab.
The hex takes that architecture and gives it a shape. Instead of connecting every MCP server you can find — which is the default behavior for anyone who discovers the MCP ecosystem — you connect six. The ones that map to your actual weekly work. The ones you'll use on a Monday morning when you're behind, not the ones that looked impressive in a demo on Saturday night.
What the Hex Is Not
The hex is not a productivity system. It doesn't have pillars, or phases, or a certification track. There is no Hex Level 2. It's a constraint applied to your specific work, and the entire value proposition is that it's small enough to actually use.
It's not a framework for frameworks. The hex doesn't organize your other systems. If you're currently running a Getting Things Done setup inside Notion with Todoist integrations and a Zettelkasten for notes — the hex doesn't layer on top of that. It replaces the impulse to build more layers. You don't need another organizational system. You need to connect six tools and use them.
It's not a course. The PDF is twenty pages. You can read it in thirty minutes and build your first hex in another sixty. If you need three months of lessons to implement a six-slot framework, the problem isn't the framework. The problem is that building the system has become a substitute for doing the work the system is supposed to support.
And it's not prescriptive about tools. The hex doesn't say "use Claude for your thinking slot and Ghost for your publishing slot and Gmail for your communication slot." It says "you have a thinking slot, a publishing slot, and a communication slot — fill them with whatever you actually use." The structure is universal. The contents are yours.
The PDF as Starting Point
The Hex Constraint PDF is a twenty-page document that walks you through the audit, the mapping, and the first build. It's genuinely useful. It's well-designed. And it is — by design — the place where most people get stuck.
Here's what happens. You download the PDF. You read the audit section — "What do you actually do every week?" — and you start filling it out. You get to the skill mapping and realize you're not sure whether your pain point is email management or the fact that your publishing workflow requires six manual steps. You start mapping skills, run out of clarity around slot four, and decide to "come back to this tomorrow." Tomorrow becomes next week. The PDF sits in your Downloads folder next to the other PDFs that were going to change everything.
This is not a design flaw. The framework is simple. Implementing it requires you to make actual decisions about your actual work — and that's where the friction lives. The PDF gives you the structure. The remaining articles in this series give you the context. What MCP actually is, why constraints work psychologically, what goes in each slot, how to build the thing in ninety minutes, and what to do after it's running. The PDF is step one. Comprehension is steps two through eight. That's what this series is for.
Who This Is For
The hex is for the person who already has too many tools and needs a filter, not another recommendation. If you're subscribed to six AI newsletters, have a bookmark folder called "Tools to Try," and spent last weekend watching automation tutorials instead of building automations — this is for you. Not because you're doing something wrong, but because the tool ecosystem is specifically designed to keep you browsing. New launches every week. New features every month. New "this changes everything" videos every day. The hex is the off-ramp.
It's also for the person who's built one or two things that actually work and wants to formalize the setup. You've got Claude doing some drafting, maybe an n8n workflow handling some scheduling, a few MCP connections you wired up manually. It works, but it's held together with duct tape and good intentions. The hex gives that duct-tape setup a shape — not by adding complexity, but by naming what you already have and cutting what you don't need.
If you've never used an AI tool beyond basic ChatGPT prompting, the hex is probably premature. Get a Claude subscription, connect one MCP server, and use it for a month. Then come back. The constraint only works when you have enough experience to know what you're constraining.
This is part of CustomClanker's Hex Explained series — you downloaded the hex, here's context.