automation-architect

What it does

A full six-phase "Agent Builder OS" that takes an automation idea from raw thought through tool selection (Zapier vs. Make vs. n8n vs. Gumloop vs. Lindy vs. Claude API vs. Openclaw) to a working, deployed system, built for one-question-at-a-time, low-overwhelm execution.

How it works

Six mandatory phases, always in order, never skipped:

• Idea Clarification: what's being built, who uses it, what changes, what pain it removes
• System Design: agents vs. linear workflow, inputs, outputs, trigger
• Tool Selection: a rule-based filter first (e.g. "multi-step workflow with complex logic → Make," "heavy reasoning or tool use → Claude API"), then adjusted for complexity, integrations, AI dependency, and scale
• Workflow Mapping: every step as trigger → action → output → next step, including error states
• Implementation: the smallest possible steps, one at a time, with checkpoints every 3-5 steps
• Optimization & Scaling: bottlenecks, cost, reliability at 10x volume, what could become agent-driven

Each phase locks its output before the next one starts.

Try it

I want to automate [what]. Walk me through this one question at a time, in order:
1. What exactly am I automating, who uses it, what's the outcome, what problem does it fix?
2. Is this agent-based or a linear workflow? What are the inputs, outputs, and trigger?
3. Given what I've described, what tool fits best (Zapier, Make, n8n, Gumloop, Lindy,
   Claude API, or Openclaw) and why?
4. Map the full workflow as trigger → action → output for every step, including what
   happens if a step fails.
5. Then walk me through implementing it, one small step at a time.

When to reach for it

Any time the plan is "connect a few apps together with some AI logic in between" and it's not obvious yet which tool should run it.