The Harness
The minimum viable harness. Why we don't sell it. Why the same Pack runs in Claude Code, Codex, n8n, Zapier, and shell scripts.
By Aaron C. Ernst · 3 min read · 2026-04-28
What you will learn
The minimum viable harness. Why we don't sell it. Why the same Pack runs in Claude Code, Codex, n8n, Zapier, and shell scripts.
blueprint
Owner guide
The harness is yours. Always was. We don't ship one, we don't host one, and we don't want the keys to one.
A harness is the place where AI actually does the work. Claude Code on your laptop. Codex in your terminal. Cursor inside your editor. n8n on a server you rent. Zapier or Make wired into a Google Sheet. A shell script your dev wrote at 11pm. If a model executes a task in it, it qualifies.
Most Bosses already have one. They haven't named it yet. The thing you've been pasting prompts into for the last six months: that is the harness. The five Zapier zaps that move leads from a form to a CRM, also the harness. The custom Cursor setup your contractor built, same category.
Here is the part the rest of the market won't tell you. The harness is where the work happens, and it's also where the data, the IP, and the chain of command live. The Boss who commands the harness owns the business. The Boss who rents it from a vendor is a tenant.
So we made a choice. We don't sell the harness. We sell the recipes the harness runs.
That's what a Pack is. A Pack is a recipe. Standing orders, prompts, schedules, integrations, guardrails. Written down, version-controlled, portable. The same Lead Rescue System runs inside Claude Code on Tuesday and inside n8n on Wednesday. Switch harnesses and the Pack still works. There's a Pack for that, and the Pack outlives the tool.
This is the opposite of how the orchestrator vendors price you. Lindy meters credits. Zapier meters tasks. Relevance meters actions plus vendor markup. Every workflow you build inside their walls is a workflow you can't take with you. We refuse to play that game. The Pack you buy from us today runs in the harness you move to next year, on the model that didn't exist when you bought it.
The metaphor we use, and the only one we use. You are the Boss. You tell the Co-pilot where to go. The harness is the Cockpit. The Pack is the standing order. You don't need to know how to fly. You need to know how to direct the Co-pilot and set the standing order, and you need a Cockpit you trust.
Most Bosses don't have one yet. They have fragments. A subscription here, a script there, a no-code workflow nobody documented. The Harness Checklist is the eight-question version of "do you actually have a Cockpit, or do you have a pile of cables."
Run the check. Then pick the harness that matches the answer. Most Bosses land on Claude Code or Codex inside a paid AI subscription they already pay for. Some land on n8n self-hosted. A handful, usually the ones with a real engineering team, land on something custom. Any of those is a fine answer. None of them is BossMode's to sell you.
Stop using AI. Start running your business on it. Own the harness. Buy the Packs.
Key takeaways
- 01The minimum viable harness. Why we don't sell it. Why the same Pack runs in Claude Code, Codex, n8n, Zapier, and shell scripts.
- 02The harness is yours.
- 03Always was.
Take the Bottleneck Check.
Sixty minutes. We map the bleed and name the Packs that stop it. Without trust, you're a bust.
Take the Bottleneck CheckKeep reading
Keep moving through the system
Best harness stacks at $5K, $7K–$15K, $100K monthly revenue
Three named stacks for three revenue tiers. Specific tools, specific monthly costs, what each tier sacrifices.
11 min read
The minimum viable harness
What a working harness needs. Three starter stacks at three budget levels. Why we don't sell the harness.
10 min read