MageTech Lead Intelligence

MageTech Lead Intelligence

Turn leads into revenue — faster and smarter with AI.

Key Capabilities

Technology Stack

System Architecture

Three clean tiers. The browser talks only to the Spring Boot API; the API owns all rules; PostgreSQL (with PGVector for AI search) is the single source of truth. Every record belongs to an organization (tenant).

Browser (Next.js SPA, port 3000)
   │  GET /api/v1/*  +  Authorization: Bearer <accessToken>
   ▼
Spring Security filter chain (stateless, CSRF off, /auth|/webhooks public)
   │  JwtAuthenticationFilter: signature + expiry + authorities (ROLE_/PERM_)
   ▼
@RestController  →  routing + DTO validation (thin controllers)
   ▼
Service layer  →  explicit permission checks + domain rules
   ▼
Spring Data JPA repositories, scoped by organization_id  →  PostgreSQL

On a 401, the frontend fetch client (lib/api.ts) transparently refreshes and retries once. The 7-step request lifecycle keeps one end-to-end round trip: browser → security → JWT → controller → service → repository → response DTO.

Business Flows

• Import pipeline: UPLOADED → PARSED → MAPPED → VALIDATED → DUPLICATE_REVIEW → CONFIRMED → COMPLETED (or FAILED / CANCELLED). Duplicate detection order: email → LinkedIn URL → company domain → company + contact.
• Lead lifecycle: IMPORTED → NEW → VALIDATED → CONTACTED → REPLIED → QUALIFIED → MEETING → PROPOSAL → NEGOTIATION → WON / LOST (ARCHIVED). Campaign send → CONTACTED; reply → REPLIED; CRM stages sync QUALIFIED / PROPOSAL / NEGOTIATION / WON / LOST back.
• Campaign pump: every 60 s scan ACTIVE campaigns → due pending enrollees (nextStepAt <= now) → send current step (SENT / FAILED / SKIPPED) → step + 1, nextStepAt = now + waitDays·86400. Missing template/email → FAILED; suppressed/empty recipient → SKIPPED.
• Reply processing: webhook / manually created reply → match outbound message → create Reply → suppress sender → CampaignLead REPLIED → Lead REPLIED → (optional) auto-create Opportunity when no open opportunity exists and auto-create flag is enabled.
• AI scoring: LLM configured → gpt-4o-mini strict JSON parse; else rules (valid email +30, website +10, employees +10, industry +10, country +10, tech +10, SEO +10, commerce tech +10) → score 0–100 → persist AiAnalysis (JSONB) → return result.
• CRM stage → lead sync: NEW 10% → — · QUALIFIED 20% → lead QUALIFIED · PROPOSAL_SENT 40% → lead PROPOSAL · NEGOTIATION 60% → lead NEGOTIATION · WON 100% → lead WON · LOST 0% → lead LOST.

Security & Multi-tenancy

• Stateless JWT (60-min access, 7-day refresh), BCrypt (strength 12) password hashing, deactivated users cannot sign in.
• RBAC enforced in services: 5 roles × 14 permission constants. Only the Owner creates/promotes Owner/Admin; no self-deactivation; the last active owner is always protected.
• Every table carries organization_id (BaseOrgEntity) — records are never shared across tenants.
• One error shape everywhere: { error, message, timestamp, status }; responses are DTOs, never entities.

Reference — Status Enums

Roles & Permissions

Permission constants: LEADS_READ · LEADS_WRITE · LEADS_DELETE · IMPORTS_READ · IMPORTS_WRITE · COMPANIES_READ · COMPANIES_WRITE · SEGMENTS_READ · SEGMENTS_WRITE · USERS_READ · USERS_WRITE · ORG_READ · ORG_WRITE · AUDIO_READ.

Pricing & Billing Model

Usage-aware workspace plans that bundle capacity — leads, AI analyses, enrichment and campaign contacts — rather than purely per-seat licensing. 14-day free trial (no credit card), add-ons available in three tiers for usage spikes (no forced upgrades).

Billing module is proposed (subscriptions + usage meter + invoices + hard/soft limit enforcement surfacing as QUOTA_EXCEEDED). Prices in INR subject to GST.