MarketWar OS
The AI-powered growth & commerce operating system. MarketWar OS enables anyone to discover
opportunities, create and market products, acquire customers, automate operations, optimise
revenue, and scale a business — from a single unified platform.
Stop guessing. Launch, test, kill, improve, and convert automatically.
The look
→ Full screenshot gallery — all 19 screens
Most tools create content. MarketWar OS diagnoses the business, rebuilds the offer, creates
campaigns, launches experiments, tracks leads, protects the budget and tells the user exactly
what to do next.
Platform modules
| Module | What it does |
|---|
| Marketing Failure Audit | Scores conversion risk, offer weakness, audience mismatch, trust and funnel leaks — the "why you got 0 customers" report |
| Executive Command Center | Live metrics bar, AI priority panel ranked by £ impact, campaign and conversation feeds |
| Campaign War Room | Every campaign carries a blunt SCALE / FIX / STOP verdict with the exact next action |
| One-Click Campaign Builder | Pick a goal → objective, audience, budget split, kill criteria and 7-day battle plan |
| Offer Builder | Volume, margin and recovery offers engineered with urgency and margin-safety checks |
| Content Factory | Strike plans where every post routes attention into channels you own |
| Video War Room | One idea → global video campaigns: generator, captions (4 modes), translation/dubbing, avatars, repurposing (1 long video → 40+ assets), brand kit, approvals |
| Landing Page Generator | Conversion pages + WhatsApp flow + 48-hour follow-up sequence designed as one system |
| WhatsApp Sales Center | Ad → WhatsApp → AI qualification → order pipeline with intent scoring |
| Customer Intelligence Vault | Contacts scored for engagement, intent, churn risk and recoverable revenue |
| Lead Recovery Engine | AI Revenue Recovery Score™ — reactivates the money sleeping in your database |
| Budget Protection | Pauses spend that produces no leads and reroutes it to winners, with a receipt |
| Competitor Spy | Threat board, exploitable gaps and ready-to-launch counter-campaigns |
| Local Domination | Google Business attack plans, community distribution, geo-offer grids |
| Revenue Intelligence | Attribution by campaign/channel, leak detection and 30-day forecasts |
| Daily Briefing | The AI Growth Strategist issues max five ranked orders per day |
The agent corps
Business Diagnosis · Customer Pain · Offer Builder · Ad Creative · Campaign Commander ·
Budget Protection · Lead Capture · Competitor Spy · Local Growth · Revenue Intelligence ·
Content Factory · AI Growth Strategist
Every agent runs under the Master Directive: zero generic info, money first, blunt
SCALE / FIX / STOP verdicts, local fidelity, structured output.
Getting started
npm install
npm run dev
Open http://localhost:3000.
The platform boots in Demo Intelligence mode — every module works with zero configuration
using deterministic simulated agent output and a demo business (Brixton Grill House).
Going live — the AI Gateway
Every agent call goes through the AI Gateway, a unified layer over three providers
with automatic failover: Anthropic Claude → OpenAI → Google Gemini. Copy
.env.example to .env.local and add any (or all) of:
AI_GATEWAY_ORDER=anthropic,openai,gemini # routing priority
ANTHROPIC_API_KEY=sk-ant-... # primary — Claude
ANTHROPIC_MODEL=claude-opus-4-8
OPENAI_API_KEY=sk-proj-... # failover — OpenAI Responses API
OPENAI_MODEL=gpt-5-mini
GEMINI_API_KEY=AI... # failover — Google Gemini
GEMINI_MODEL=gemini-2.5-flash
The gateway tries providers in order, retries transient errors (429/5xx) with
exponential backoff per provider, and fails over to the next configured provider.
GET /api/gateway reports the live routing status (keys are never exposed).
Never commit .env.local or real keys — .gitignore already excludes env files.
Architecture
src/
├── app/
│ ├── page.tsx # Marketing landing page
│ ├── how-it-works/ # 7-phase mission protocol
│ ├── onboarding/ # Business Brain intake (4 steps)
│ ├── dashboard/ # The operating system
│ │ ├── page.tsx # Executive Command Center
│ │ ├── audit/ # Marketing Failure Audit report
│ │ ├── war-room/ # Live campaign grid + verdicts
│ │ ├── campaigns/ # One-Click Campaign Builder
│ │ ├── offers/ content/ landing-pages/
│ │ ├── whatsapp/ customers/ recovery/
│ │ └── revenue/ budget/ competitors/ local/ briefing/
│ └── api/
│ ├── agents/[agentId]/ # Agent execution endpoint (live or demo)
│ ├── gateway/ # AI Gateway status endpoint
│ └── audit/ # Deterministic failure-audit scoring engine
├── components/ # UI kit, sidebar, AgentRunner harness, chart kit
└── lib/
├── ai/agents.ts # Agent registry: prompts + demo outputs
├── ai/gateway.ts # AI Gateway: Claude/OpenAI/Gemini adapters + failover
├── ai/provider.ts # Agent runtime on top of the gateway + demo fallback
├── ai/audit.ts # Failure-audit scoring engine
├── data/demo.ts # Demo Intelligence dataset
└── types.ts # Domain types
Production architecture & deployment
The adopted production standard is Hostinger (domain) → Cloudflare (edge) →
Vercel (frontend) → Firebase (Auth, Firestore, Storage, Functions, Cloud Run),
with Stripe for billing and the AI Gateway for model providers — specified in
docs/PRODUCTION-ARCHITECTURE.md, with the
step-by-step go-live runbook in docs/DEPLOYMENT.md.
Firebase wiring ships in src/frontend/firebase-client.ts + src/backend/firebase-admin.ts and activates via env vars;
firestore.rules / storage.rules are the deny-by-default baselines.
AI-OS engineering blueprint
The complete production-grade specification for the AI Infrastructure Operating
System build-out — executive vision, market-gap review, user command centres, the
full agent workforce, module specs, the BitriPay payment gateway door, connector
ecosystem, architecture, database schema, API spec, monetisation, security and the
phased roadmap — lives in docs/. The original v3.0 source
specification is preserved unmodified in
docs/reference/.
Scripts
npm run dev — development servernpm run build — production buildnpm run start — serve the production buildnpm run typecheck — TypeScript check
