Our stack
We preach owning the machine. Here’s ours.
StackSwap replaces rented GTM sprawl with one engine you own: a buyer model trained on your data, scoring logic you can inspect, and outcomes that compound instead of expiring at renewal. This page is the receipt — the product surfaces already shipped, the data layer underneath, and the rented rails we use to move fast without giving up control.
Owned where it matters. Rented where it should be.
Why publish this?
What this page proves
Most GTM vendors ask you to trust a black box: data goes in, a number comes out, and the scoring logic stays locked in someone else’s account. StackSwap’s bet is the opposite. The model, the scoring logic, the data flows, and the recommendations should be inspectable, portable, and explainable. So we built ours that way — and this page documents it in the open.
Built on a real production stack, not a mockup
Every card below links to a long-form write-up of something that exists in the codebase today — runtimes, schemas, and product surfaces, not slideware.
The recommendation engine and knowledge base are already shipped
The free tools, comparison pages, and StackScan flow run in production now. The StackSwap OS product itself is still pre-beta (closed sandbox beta opens July 1, 2026).
The same data model already powers audits, reports, and comparisons
One canonical layer feeds the StackScan audit, the generated GTM report, and the premium comparison pages — so the numbers stay consistent across every surface.
The machine behind the free tools becomes the intelligence layer in StackSwap OS
The engine that scores and explains public stacks is the same engine that becomes your owned buyer model inside StackSwap OS — it compounds instead of resetting at renewal.
Owned where it matters: data model, scoring logic, recommendation engine, product surfaces, and deployment code.
Rented where it should be: payments, hosting, LLM providers, and infrastructure rails.
Own the model
The machine we own
Product surfaces already live — not roadmap claims.
StackScan Flow
URL-first intake, Supabase-backed lead capture, and unified GTM Engine results route.
Read →Recommendation Engine
7 stack blueprints, 6 fit rules, company-specific stack recommendations via Claude.
Read →Tool Capability Map
16 tools with functions, strengths, limitations, replacement rules, fit logic.
Read →AI-Native Gap Metric
Measures legacy exposure as a percentage. Color-coded. Benchmarked against industry.
Read →GTM Report Template
10-page dark-mode GTM strategy HTML template and TypeScript generator wired for PDF export.
Read →Premium Comparison Pages
13 /compare/[slug] pages with Article + FAQPage JSON-LD, structured verdicts, and editorial tuned for LLM-SEO.
Read →Knowledge Base Engine
Comparison hub, category pages, 60+ articles. FAQ schema + canonical URLs drive LLM-SEO across the entire KB.
Read →Cross-Tool Data Flow
StackScan, StackBuilder, StackEnrich, and GTM Strategy pass context via URL params + sessionStorage.
Read →Company URL Inference
Server-side company lookup API that infers name, industry, positioning, and competitors from a single URL.
Read →Sortable Plan Table
Tri-state column sort with affordance arrows. Shared component between paid, free preview, and pre-paywall surfaces.
Read →Operating Profile Inference
Unified profile object for demo + real users, powering tailored executive summary and benchmark credibility lines.
Read →StackScan Data Adapter
Adapter that maps StackScan / StackBuild analysis and form context into a single GTM strategy report model.
Read →Report Checkout + Unlock Flow
Stripe Checkout for one-time $99 report unlocks, webhook verification, and report-level premium state.
Read →Own the model
The data layer underneath
The systems that make recommendations consistent, scored, and auditable.
Supabase Data Layer
21-column schema. RLS policies. Source tagging. Organic intelligence views.
The shared store behind every scan, report, and benchmark, so recommendations stay consistent across surfaces.
Read →1,120-Tool Database
Per-seat + flat-rate pricing. Team multipliers. Industry benchmarks. 837 priced.
The priced catalog behind overlap detection, swap suggestions, and cost math.
Read →11-Source Weighted Pricing Council
11 independent SaaS pricing datasets merged with confidence weighting behind every tool cost estimate.
Combines multiple pricing sources with confidence weighting, so estimates aren’t based on one stale number.
Read →Canonical Money Architecture
One-number-across-surfaces rule: hero, plan table, stack transformation, and CTA all read the same canonical source.
One source of truth for pricing math, so different pages can’t show different cost estimates.
Read →Leads Schema v2
Extended Supabase leads table for StackScan: GTM fields, analysis scores, spend estimates, source tagging.
One structured record per submission, so analytics and reporting stay clean and segmentable.
Read →Team-Size Precision Pipeline
Precise numeric headcount threads through scanStack end-to-end. No more lossy band round-trip between intake and scan.
Carries your exact headcount through the whole scan, so a 15-person team and a 45-person team get different math.
Read →Circular Redundancy Safeguards
Dedup logic blocks "remove A because B / remove B because A" rationales from ever reaching the execution timeline.
Prevents the engine from recommending circular or duplicate replacement logic.
Read →Recommendation Normalization Layer
Deterministic validation pass that resolves conflicting KEEP/REPLACE/REMOVE/ADD outcomes before render.
Resolves conflicting keep/replace/remove calls before you see them, so the plan stays internally consistent.
Read →StackScan Fuzz Harness
500-scan invariant sweeps across 12 rules. Catches pricing drift, circular logic, and phantom savings before users do.
Runs hundreds of synthetic stacks through the engine before launch, so the savings math is proven consistent before you see it.
Read →Internal BI Dashboard
Real vs seed filtering. Funnel analysis. Password-gated.
Separates real traffic from seed data, so internal metrics and public claims stay honest.
Read →Platform.js
Shared nav + footer, metrics, Supabase, count-up animations. Pufferfish.
Shared navigation, footer, and data wiring, so every surface behaves like one product, not a pile of apps.
Read →main.css Design System
Single CSS source of truth. Cards, buttons, inputs, metrics, pills, grids. All 14 pages.
A single styling source of truth, so every page stays visually consistent as new ones ship.
Read →Rent the rails
The rails we rent
Replaceable infrastructure we rent to stay fast, portable, and cheap to run.
Vercel Edge + Serverless
Preview deploys per PR. Edge + Node runtimes. Cron + background jobs for StackScan work.
Read →Next.js Hybrid Router
Pages + App Router in one codebase. Marketing on Pages, StackScan results on App.
Two routing modes in one codebase, so we can ship new surfaces fast without rewriting working pages.
Read →TypeScript End-to-End
Strict TS across UI, API, and server code. Shared types for reports, leads, intake.
The same typed contract runs from the database to the screen, so a schema change can’t silently break a report.
Read →Stripe Payment Rail
Checkout sessions, webhooks, fulfillment state, idempotency — report-bound, no accounts.
Read →LLM Multi-Provider
OpenAI + Claude behind a common surface. Structured outputs, token-usage logged per report.
Read →Edge API Proxy
Vercel serverless. CORS handling. Server-side API keys.
Keeps API keys and heavy logic on the server, so secrets never ship to your browser.
Read →“The operators who learn what makes their buyers convert — and compound that knowledge every month — are going to eat.”