I get asked about my stack every week. Here's the honest answer — no affiliate links, no "best tools of 2026" padding. Just the tools that are running in production right now, what they cost, and why I chose them.
The Selection Criteria
Before we get into the list, here's how I evaluate tools:
- API-first: If I can't automate it, I don't use it. Pretty dashboards are worthless if the system can't call them.
- Cost-per-unit: Not monthly price — cost per actual operation. A $50/mo tool that handles 100K operations is cheaper than a free tool that takes 10 minutes of manual work each time.
- Failure modes: What happens when it breaks? Tools that fail silently are dangerous. I need errors I can catch and handle.
- Replacement cost: How locked in am I? If the tool doubles its price tomorrow, can I switch in a weekend?
1. AI Engine — Claude (Anthropic)
Cost: $30–$80/mo depending on volume
Why: Best instruction-following in the market. When you're building systems that need to reliably produce structured output — JSON, specific formats, categorized content — Claude is the most consistent. GPT-4 is close, but Claude's system prompts are more deterministic.
The real reason: For operator work (writing, analysis, code generation), Claude makes fewer "creative interpretation" errors. When I say "write exactly 5 tweets in this format," I get exactly 5 tweets in that format. Every time.
2. Orchestration — OpenClaw
Cost: $0–$25/mo
Why: The agent runtime. This is what turns "AI that answers questions" into "AI that runs systems." Handles scheduling, tool access, memory, multi-step workflows. It's the difference between having a chatbot and having an operator.
Key feature: Persistent memory across sessions. My AI doesn't start fresh every conversation — it knows what it did yesterday, what's scheduled, and what to check on.
3. Hosting — Vercel
Cost: $0–$20/mo
Why: Next.js deployment in one push. Zero config. Edge functions for anything dynamic. The free tier handles more traffic than most new products will see in their first year.
Alternative: Railway ($5/mo) if you need persistent servers, background jobs, or databases without the serverless constraints.
4. Email — Resend
Cost: $0–$20/mo
Why: API-first email. No drag-and-drop editors, no template builders — just send an HTTP request with HTML and it arrives. For automated email sequences (welcome series, post-purchase drips, testimonial requests), this is all you need.
The operator advantage: Every email in my system is triggered by code, not by clicking "send campaign." AI drafts it, the system schedules it, Resend delivers it. I don't touch it.
5. Payments — Stripe
Cost: 2.9% + $0.30 per transaction
Why: Not even a choice. Stripe's API is the gold standard. Payment links for quick launches, webhooks for post-purchase automation, and the dashboard for when you need to debug.
Pro tip: Start with Payment Links, not a custom checkout. You can build the beautiful checkout page later. On day one, a Stripe Payment Link works perfectly and takes 30 seconds to set up.
6. Analytics — Custom + Plausible
Cost: $0–$9/mo
Why: Plausible for web analytics (privacy-friendly, lightweight, no cookie banners needed). Custom scripts for business metrics — tweet performance, email open rates, conversion funnels.
The key insight: Separate your vanity metrics from your revenue metrics. Plausible shows you traffic. Your custom system shows you money. Only one of those matters.
7. Content Distribution — xurl (X/Twitter API)
Cost: $0 (API access)
Why: Direct API access to post, engage, and measure on X. No third-party scheduling tools with their own markup. The AI writes the tweets, the scheduler picks the time, xurl does the posting. Clean.
Why not Buffer/Hootsuite: They add a layer between you and the platform that you don't control. For an operator, every layer you don't control is a liability.
The Total Stack Cost
Minimum viable: $31/month.
Production grade: $100–$160/month.
For that, you get: AI-generated content, automated email sequences, payment processing, web hosting, analytics, and social distribution. All running without daily intervention.
What I Didn't Include (And Why)
No CRM. A spreadsheet or JSON file handles customer tracking until you have 500+ customers. Adding Salesforce at 20 customers is like buying a forklift for your apartment.
No project management tool. The AI agent tracks its own tasks. Adding Notion or Linear on top creates two sources of truth, and two sources of truth means zero sources of truth.
No design tools. AI generates OG images, email templates, and basic graphics. For anything more complex, I'll hire a designer for a one-time project. Monthly Figma subscriptions for occasional design work is money on fire.
Start With Three
If you're building from scratch, start with three tools: AI engine, hosting, and payments. That's enough to have a product, a website, and a way to charge for it. Add orchestration when you're tired of doing things manually. Add analytics when you have enough data to analyze. Add email when you have someone to email.
The operator stack isn't about having every tool. It's about having the right tools, integrated tightly, running without you.
Want the full tool comparison? The Operator Playbook includes Appendix A: The Operator's Toolkit — detailed comparisons across 5 categories with costs, pros/cons, and specific recommendations for different business sizes.