BrandArc
AI-powered brand identity generator — strategy, logos, colors, and typography in under 30 minutes
Mar 2026 – Present
Visit Live4 Chained
AI Agents
391 (1,309 assertions)
Tests
~30 min → concurrent
Generation Time
₹1,999
Price Point
The Challenge
Indian startups face a harsh branding gap: agencies charge ₹50K+ and take 4–6 weeks, freelancers are unreliable, and DIY tools like Canva don't produce cohesive brand identities. Most founders launch with inconsistent branding — mismatched colors, generic logos, no brand strategy — which hurts credibility with investors and customers.
BrandArc automates the entire brand identity process using a chained AI pipeline. A founder answers 5 questions about their business, and the system generates a professional brand kit: positioning strategy, 3 logo variants, a 5-color palette, and typography pairing — all cohesive and industry-optimized.
My Role
Solo founder and full-stack developer — I designed, built, and ship the entire product end-to-end:
- Product strategy and market positioning
- System architecture and AI pipeline design
- Frontend (React 19 + Inertia.js v2)
- Backend (Laravel 12 + PostgreSQL)
- Payments integration (Razorpay)
- Infrastructure and deployment
- Testing (391 tests, 1,309 assertions)
- SEO and analytics
Technical Decisions
Why chained AI agents?
A brand identity isn't random outputs stitched together — strategy must inform the color palette, and the palette must inform logo generation. Chaining 4 agents ensures every element is cohesive. The strategy agent's output (industry, tone, personality) feeds directly into palette and typography agents, which then feed the logo agent.
Why Inertia.js v2?
Inertia gives a full SPA experience while keeping Laravel's routing, middleware, and controllers intact — no separate API layer to build or maintain. For a solo founder, this cuts development time significantly without sacrificing UX quality.
Why Razorpay?
The Indian market needs UPI, cards, net banking, and wallets — Stripe doesn't support UPI. Razorpay's UPI flow is async and webhook-based, requiring polling on the frontend and idempotent verification on the backend to handle network interruptions during payment.
AI Pipeline Architecture
┌──────────────────────────────────────────────────────────────┐
│ BRAND GENERATION PIPELINE │
│ │
│ 5-Step Wizard ──→ CreateBrand Action ──→ GenerateBrand Job │
│ │ │
│ ┌─────────┴──────────┐ │
│ ▼ │ │
│ BrandStrategyAgent │ │
│ (GPT-4o-mini) │ │
│ │ │ │
│ ┌─────────┴─────────┐ │ │
│ ▼ ▼ │ │
│ BrandPaletteAgent BrandTypographyAgent │
│ (concurrent) (concurrent) │ │
│ └─────────┬─────────┘ │ │
│ ▼ │ │
│ BrandLogoAgent │ │
│ (DALL-E 3) │ │
│ │ │ │
│ ▼ │ │
│ ProcessLogoFormats │ │
│ (PNG → transparent → SVG → PDF) │ │
│ │ │ │
│ ▼ │ │
│ Notify User (email) │ │
│ │ │
│ Supports: Cancel mid-generation, Retry on failure, │ │
│ Concurrent palette+typography, 3 retry │ │
│ attempts with exponential backoff │ │
└──────────────────────────────────────────────────────────────┘
The pipeline runs asynchronously via Laravel queues. Each agent has structured JSON output schemas enforced by the AI SDK, and cancellation checkpoints between each step prevent half-generated brands from corrupting state.
Freemium & Payments
Tier Breakdown
- Free tier — 1 brand generation with watermarked logos (512px), no downloads
- Pro tier — ₹1,999 one-time payment for 3 regenerations, high-res logos, ZIP/PDF downloads, and SVG export
Upgrade Flow
Existing free brands can be upgraded to Pro for ₹499 — watermarks are removed, logos are re-generated at full resolution, and all download formats unlock.
Payment Security
Razorpay signature verification on every webhook with replay attack prevention. UPI payments are fully async — the frontend polls for confirmation while the backend handles idempotent webhook processing.
Engineering Highlights
- 391 tests, 1,309 assertions — comprehensive coverage with Pest v4
- 22 Action classes — single-purpose pattern keeping controllers thin
- PHPStan Level 5 (bleeding edge) + Pint, ESLint, and Prettier
- 7 JSON-LD schemas — dynamic sitemap, robots.txt, and OG image generation
- 26 Inertia/React pages — skeleton loaders, animations, and dark mode throughout
- 10 Eloquent models — with factories, scopes, and authorization policies
- 2FA/TOTP authentication — login, register, email verify, password reset, profile management
What I Learned
-
Chaining AI agents requires careful state management — Each agent's output feeds the next, so partial failures mid-pipeline need graceful handling. I built cancellation checkpoints between each agent and used database transactions to prevent half-generated brands.
-
Freemium pricing for AI products is tricky — AI generation has real per-unit cost (OpenAI API calls). The free tier must be constrained enough to drive upgrades but generous enough to demonstrate value. Watermarking + resolution limits achieved this balance.
-
Concurrent AI calls cut latency ~40% — Running palette and typography agents in parallel (they only depend on strategy, not each other) reduced total generation time significantly.
-
Indian payment integration has unique challenges — Razorpay's UPI flow is async (webhook-based), requiring polling on the frontend and idempotent verification on the backend to handle network interruptions during payment.
-
Building product is the easy part — A well-engineered product with zero users is worth nothing. The revised roadmap front-loads analytics, SEO, and retention mechanics before adding more features.