Hex Configurations by Role: Writer, Developer, Creator, Consultant

The hex is a constraint, not a prescription. Your six tools depend entirely on what you actually do. A writer's hex looks nothing like a developer's hex, which looks nothing like a video creator's hex. What they share is the structure — six tools, wired together, used daily with depth. What differs is what fills the slots.

This article walks through four hex configurations for four common roles. These are not "the best tools for writers" or "the recommended stack for developers." They're examples of how the hex constraint plays out when applied to real workflows. Your hex will be different because your work is different. The value here is seeing the decision process, not copying the output.

The Writer's Hex

The writer publishes. Articles, newsletters, blog posts, scripts — the output is text that goes somewhere for someone to read. The writer's workflow is: research, draft, edit, publish, distribute. Every tool in the hex needs to serve one of those five stages.

Slot one: Claude. The thinking and drafting tool. This is where the writer does first drafts, outlines, research synthesis, and editorial passes. Claude fills this slot over GPT for most writers because the writing quality at longer lengths is meaningfully better — Claude maintains voice consistency across 2,000-word pieces in a way that GPT still struggles with as of early 2026. The writer's system prompt defines their voice, their structure preferences, and their publication format. This is the most-used tool in the hex by a wide margin.

Slot two: Ghost (or whatever CMS they publish on). The publishing tool. Ghost earns the slot for independent publishers because it handles content, email, and membership in one platform, which means one tool covers two workflow stages — publish and distribute. If the writer publishes on WordPress, Substack, or something else, that tool fills this slot instead. What matters is that it has MCP support, so Claude can create drafts directly in the CMS without the writer manually copying markdown between tools.

Slot three: Gmail (via MCP). The communication tool. Email is where the writer coordinates with editors, sources, clients, and readers. Having Gmail connected through MCP means Claude can read emails for context, draft responses, and handle the communication overhead that writers often procrastinate on. The alternative — switching to a browser tab, reading the email, switching back, explaining the context to Claude — adds ten minutes of friction per email interaction.

Slot four: Google Drive or local file system (via MCP). The knowledge tool. This is where research notes, source material, style guides, and reference documents live. The writer needs Claude to be able to search their files, pull relevant notes, and reference past work. Whether this is Google Drive, Dropbox, Obsidian through a local MCP connector, or just a well-organized folder structure on the local file system depends on the writer's existing habits. The critical thing is that Claude can access it.

Slot five: one image generation tool. The creation tool. Most published writing needs images — feature images, in-article illustrations, social thumbnails. The writer needs one tool for this, not three. Which tool depends on what kind of images the writer needs and — critically — whether it has API or MCP access. A tool that produces beautiful images but requires the writer to download, upload, and manually place them is less useful than a tool that produces good-enough images that Claude can generate and place directly.

Slot six: n8n or equivalent. The automation tool. This handles the recurring parts of the publishing workflow — scheduling posts, cross-posting excerpts, formatting email digests, backing up content. The automation slot is the least used daily but the most valuable weekly, because it handles the tasks the writer would otherwise skip or do inconsistently.

What the writer cut: SEO tools (Claude handles keyword research well enough for most independent publishers), social media schedulers (the automation tool covers this), dedicated grammar checkers (Claude's editing is strong enough), and the second or third LLM they were using "for different tasks" (one LLM, used with depth, outperforms two LLMs used shallowly).

The Developer's Hex

The developer ships code. The workflow is: plan, write, test, deploy, maintain. The hex serves this pipeline.

Slot one: Claude Code. The development tool. Claude Code runs in the terminal, reads codebases, writes and edits code across files, runs tests, and handles git. It fills the thinking tool slot and the creation tool slot simultaneously — the developer's primary creative output is code, and Claude Code is both the reasoning partner and the production tool. This is the most used tool by hours per day.

Slot two: GitHub (via MCP or CLI integration). The publishing and collaboration tool. Code goes to GitHub. PRs, issues, code review, CI/CD triggers — it all flows through the repository. Having GitHub connected means Claude Code can create branches, commit, push, and draft PRs without the developer switching to a browser. For solo developers, this integration is convenient. For team developers, it's essential.

Slot three: the deployment platform. Vercel, Railway, Fly.io, a VPS with SSH — wherever the code runs in production. This slot is about getting code from the repository to the place where users interact with it. The ideal is a platform where Claude can trigger deployments and check status. In practice, this is often handled by CI/CD connected to GitHub, making slots two and three partially overlapping. That's fine — it means the developer effectively has a spare slot.

Slot four: the project management or communication tool. Linear, Notion, or even just a structured set of markdown files. Somewhere the developer tracks what needs to be built, what's in progress, and what's done. For solo developers, this can be minimal — a local TODO file that Claude can read. For developers on teams, it needs to be the shared tool the team uses. The key is that Claude can read and update it, so the developer can say "what's the next priority" and get an answer from the actual task list, not from Claude's guess.

Slot five: a browser or testing tool. Claude's ability to interact with web interfaces through MCP — checking a deployed app, reading documentation, verifying that a change works in the browser — fills a slot that developers traditionally handled by switching between terminal and browser. For frontend developers, this is essential. For backend developers, it might be replaced with a database tool or API testing interface.

Slot six: documentation or knowledge tool. Where the developer stores architectural decisions, API references, project context, and learning notes. For developers who work on multiple projects, this is the tool that prevents "wait, how does this project work again" on Monday morning. Claude's ability to search project documentation and context files before making changes is the difference between code that fits the existing architecture and code that ignores it.

What the developer cut: the second code assistant (no more running Cursor and Claude Code and Copilot simultaneously), the local LLM they were running "for privacy" but barely using, the monitoring stack they built for their monitoring stack, and the three automation tools they were evaluating for CI/CD when one would have been enough.

The Creator's Hex

The creator makes media — video, audio, images, or some combination. YouTube, podcasts, social content. The workflow is: ideate, produce, edit, publish, promote.

Slot one: Claude. The thinking and scripting tool. For creators, the LLM handles ideation, script writing, show notes, descriptions, titles, and the text layer of content creation. A YouTuber's Claude prompt looks different from a writer's — it's focused on hook structures, visual cues, and pacing rather than long-form prose. A podcaster's prompt emphasizes interview prep, question generation, and post-production copy.

Slot two: the primary production tool. For video creators, this is their editing suite — DaVinci Resolve, Premiere, CapCut, whatever they edit in. For podcasters, it's their DAW — Descript, Audacity, Logic. For visual creators, it's their image tool. This is the tool where the primary creative output is assembled. It's often not an AI tool at all, and that's fine. A hex slot doesn't have to be AI-powered — it has to be essential to the workflow.

Slot three: one AI media tool. This is where AI enters the production pipeline. For video creators, it might be an image generation tool for thumbnails and B-roll. For podcasters, it might be a TTS or voice tool. For visual creators, it might be an AI editing or upscaling tool. The constraint here is one — not one per media type, but one total. The creator picks the AI media capability that adds the most value and skips the rest.

Slot four: the publishing platform. YouTube Studio, Spotify for Podcasters, Instagram — wherever the content goes live. The value of having this connected (where possible) is that Claude can handle the metadata layer — titles, descriptions, tags, scheduling — while the creator focuses on the content itself.

Slot five: the communication and community tool. Email, Discord, community platform — wherever the creator interacts with their audience. Creators often underinvest in this slot because audience communication feels like overhead rather than creation. But the creators who maintain audience relationships produce content that resonates better, and having Claude handle the communication logistics frees time for the creative work.

Slot six: the scheduling and automation tool. Content calendars, cross-posting, social media scheduling, RSS distribution. The recurring operational tasks that keep the content machine running between creative sessions.

What the creator cut: the three separate AI tools they were using for thumbnails, B-roll, and background music (one AI media tool, not three), the analytics dashboard they checked daily but never acted on (check monthly in the quarterly review instead), and the social media AI tools that were generating posts nobody engaged with (manual posts from the creator performed better than AI-generated posts about the creator's content).

The Consultant's Hex

The consultant advises clients on AI tool adoption and implementation. The workflow is: research, advise, implement, deliver, follow up. The hex serves the consulting practice, not the client's tools.

Slot one: Claude. The thinking, research, and deliverable tool. The consultant uses Claude for client research, proposal drafting, deliverable creation, and staying current on the AI tool landscape. The system prompt is longer than most — it includes the consultant's positioning, their methodology, and their standard deliverable formats. Claude functions as the consultant's back office.

Slot two: the presentation and deliverable tool. Google Slides, Notion, or whatever format the consultant delivers in. Client recommendations need to look professional and be shareable. Having this connected means Claude can create and populate deliverable templates directly, rather than the consultant manually building each deck or document.

Slot three: Gmail (via MCP). Client communication. Consultants live in email — proposals, scheduling, follow-ups, deliverable transmission. Having Claude manage the email layer means the consultant spends time on advisory work, not on writing "just following up" messages.

Slot four: Google Calendar (via MCP). The scheduling tool. Client calls, implementation sessions, review meetings. Having calendar connected means Claude can check availability, draft meeting agendas, and prep call notes — the operational overhead of a consulting practice that most consultants handle manually and poorly.

Slot five: Google Drive or equivalent. The knowledge base. Client files, project history, reference materials, the consultant's own testing notes on tools they've evaluated. This is where institutional memory lives. A consultant who can say "I tested that tool for Client X last quarter, here's what happened" — because Claude can pull up the notes — is more credible than one relying on memory.

Slot six: one testing environment. A sandbox where the consultant can test tools before recommending them. This might be a VPS, a local development environment, or simply a structured set of free accounts the consultant maintains for evaluation purposes. The key constraint: the testing environment is used during scheduled testing windows (two hours per week, for example), not as an ongoing exploration hobby. The consultant tests deliberately, documents results, and moves on.

What the consultant cut: the twelve AI tool subscriptions they were maintaining "to stay current" (scheduled testing replaces ongoing subscriptions), the CRM they were paying for but never updating (email and calendar cover client management at the solo consultant scale), and the content creation tools they thought they needed for marketing (Claude handles the text marketing, and the consultant's credibility comes from being right about tools, not from having a polished social media presence).

The Meta-Pattern

Across all four hexes, a pattern emerges. The LLM — Claude, in all four examples, though it could be another — occupies the thinking slot in every configuration. This is not an accident. The LLM is the orchestration layer, the tool that coordinates the others through MCP. It's the hub of the hex, and the other five tools are the spokes. The choice of LLM determines what tools can connect and how, which is why it's the most important decision in the hex and the one that should be made first.

The other consistent pattern: every hex has at most one AI media creation tool. Not three, not five — one. The instinct to collect AI creation tools across every media type is strong, but the hex reveals that most people only produce one kind of media regularly. The writer produces text. The developer produces code. The creator produces video or audio. Collecting image, video, audio, and text generation tools "just in case" is collecting, not creating.

Your hex will look different from all four of these. The point isn't to copy a configuration — it's to see the constraint in action and recognize the decision patterns. Pick the six that serve your actual work, not the six that make you feel like you have the most coverage.

This article is part of The Hex System series at CustomClanker.

Previous: The Quarterly Review | Next: The Cost of Tool Sprawl