My Apps — Chetan Sroay

01 — Why I Build

I build for myself first.
Three live products. One engineer.

Most engineers talk about what agents could do. These three products and the platform behind them were conceived, designed, built and deployed by one person, by directing AI coding agents and then verifying what they produced. The interesting part is not that the code got written; it is knowing whether it is right.

Each product covers ground that a stack of separate tools normally covers. Built with the stack I actually use: Claude Code and MCP servers to direct the agents, then TypeScript, Next.js, React, Supabase with Row Level Security, pgvector, and five LLM providers behind one router.

founder@technobelieve ~ zsh

$ ls products/

forge/ kompense/ marketingsohigh/ believele/

$ cat stats.json

{

"ai_agents": 24,

"surfaces": ["web", "android", "chrome", "wordpress"],

"public_listings": ["play store", "chrome web store", "wordpress.org"],

"verification": "multi-agent review + fail-closed gates",

"llm_providers": 5,

"platforms_published_to": 40

}

$ echo $COMPANY

Techno Believe Solutions Ltd

Forge

AI Agent Orchestration Platform

The AI workforce that runs the other three companies. 24 autonomous AI agents, cron-scheduled around the clock, coordinated through a purpose-built MCP server — drafting content, outreach and support work. Anything high-stakes, and anything it cannot verify against a bank of checked facts, waits for my approval.

Autonomous Agents

Scheduled Jobs

Products It Runs

Architecture

Next.js 16 + React 19 Gemini + Claude SDKs Model Context Protocol Supabase SSR Zod-Validated Outputs Vercel Cron Approval-Gated Actions TypeScript

01 Role-Specialized Agents Strategy, content, distribution, outreach, analytics, and compliance — each agent has a defined role, tiered model access, and few-shot examples for consistent output.

02 MCP Server Exposes live resources and tools over the Model Context Protocol, so agents — and I — can query and act on real production state, not a mock.

03 Human-in-the-Loop Approvals Every outbound action — content, replies, outreach — routes through an approve / reject / defer / edit queue, with learnings recorded back into the system.

04 Schema-Validated Outputs Every agent response is Zod-validated with automatic retry on schema failure — structured, reliable output instead of free-form LLM text.

05 Cron Autonomy Scheduled jobs run content, metrics rollups, and market-watch tasks on a fixed cadence, independent of anyone opening the dashboard.

06 Multi-Tenant by Design One platform, three brand voices. Forge runs Kompense, MarketingSoHigh, and Believele from a shared core without blending their identities.

Ask Me How It Works →

Private infrastructure — not a public product

Kompense

AI-Powered Competitive Intelligence

Competitive intelligence for B2B SaaS product-marketing teams, and the agencies who serve them. For teams who cannot justify $30K+ a year on Klue or Crayon. An AI analyst answers questions from live competitor data across pricing, ads, content, social, reviews and hiring, and agencies can resell it under their own brand.

Tools In The Analyst Agent

23,671

Insights Collected

101

Competitors Tracked

Architecture

19-Tool Agent pgvector + HNSW Shared Intelligence Layer React + Tailwind Supabase Row-Level Security Gemini White-Label Ready

01 Real-Time Monitoring Automated collection across competitor websites, pricing pages, ads, content, social, reviews and hiring, with freshness set per module and scrape frequency rising with the number of subscribers.

02 AI Insight Engine A shared intelligence layer: a competitor is collected once and reused by every account tracking it, so data cost per account falls as the platform grows rather than rising.

03 Smart Alerts AI-prioritized alerts ranked by business impact severity — Critical, High, Medium. Never miss what matters.

04 AI Chat Assistant Natural language queries about your competitive landscape. Ask anything, get strategic intelligence instantly.

05 Export Engine Board-ready reports in PDF, Word, CSV. Custom report builder with strategic recommendations.

06 Social Intelligence Review tracking, social signals, trend analytics with historical visualization. Full market awareness.

Live Platform Preview

Kompense - AI Competitive Intelligence Platform

Visit Kompense →

Live at kompense.com with a public demo — early-stage, one confirmed external account

MarketingSoHigh

AI Organic Marketing Automation

SEO, blog publishing, social distribution and email outreach in one place. For small businesses and agencies priced out of a $5–7K-a-month stack of separate tools, with your own AI keys if you prefer. It writes and publishes the blogs of four of my own websites, which is where the numbers below come from.

727

Articles Published

82,000+

Google Impressions Since May

30 → 17

Average Google Position

Architecture

Plugin On WordPress.org 5 LLM Providers 40 Platforms Multi-Tenant Supabase BYOK Support WordPress Plugin White-Label

01 Content Engine A daily blog engine with topic and bottom-of-funnel engines, keyword demand from DataForSEO and Search Console, competitor-gap mining and content-decay detection with an automatic refresh queue.

02 SEO Suite Keyword discovery, topic clustering, SERP tracking, site audits, competitor gap analysis, and backlink research.

03 Outreach Engine AI-personalized LinkedIn DMs and cold emails. Apollo.io enrichment, email warmup, smart sequences with icebreakers.

04 22-Platform Distribution LinkedIn, Reddit, Medium, Dev.to, Bluesky, Ghost, WordPress, and 15+ more. One click, everywhere.

05 Analytics & OKRs AI strategy generation, health score (0–100), per-channel breakdowns, PDF/CSV reporting, goal tracking.

06 Lead Pipeline Scoring, forecasting, Kanban pipeline, unified inbox. From content to conversion in one platform.

Live Platform Preview

MarketingSoHigh - AI Marketing Automation Platform

Visit MarketingSoHigh →

Covers the ground a separate SEO, publishing, social and outreach stack normally covers — clicks are still the weak number, and the open problem

Believele

AI Career Intelligence Platform

Stop applying blindly. Start applying intelligently. Analyzes every job against your profile, generates ATS-optimized documents, and builds data-driven strategies — turning weeks of guesswork into minutes of precision.

Clients, One Codebase

Independent Checks Per Document

103

Job Sites The Extension Works On

Architecture

Gemini (Provider-Agnostic) Supabase + RLS Web App Android App Chrome Extension ATS Optimization NLP Analysis

01 Job Analysis Paste any listing. Get match scores, skill gap assessments, and keyword priorities instantly. Know your fit before you apply.

02 CV Generator ATS-optimized resumes tailored to each role. Restructures experience, highlights achievements, calibrates keywords.

03 Cover Letters Mirrors company tone, addresses specific requirements, positions your experience as the solution to their hiring need.

04 Career Strategy Interview prep, networking strategy, salary benchmarks, and a gap-closing action roadmap. End-to-end career intelligence.

05 Bev — AI Assistant Context-aware chatbot for job search guidance, career transitions, and professional development. Your personal career advisor.

06 Multi-Platform Web app, native Android, and Chrome extension. Apply intelligently from anywhere — desktop or mobile.

Live Platform Preview

Believele - AI Career Intelligence Platform

Believele Dashboard - Job Analysis & CV Generation

Visit Believele →

Available on: Web + Android + Chrome Extension

06 — What This Proves

Full-Stack Ownership

Design, frontend, backend, database, AI integration, deployment — one person, end-to-end. No team. No handoffs. Complete systems from zero to production.

Production at Scale

Shipped, not prototyped: on Google Play, in the Chrome Web Store, on WordPress.org and on the open web. Store review, versioned releases, over-the-air updates and permission policy are part of the work. Early-stage, and the throughput figures here are overwhelmingly my own usage.

AI-Native Architecture

Verification as architecture: independent reviewer agents attack a change before it merges, critical fixes ship with a test proven to fail without them, published content passes a fail-closed fact gate, and every document a user receives is checked by something other than the model that wrote it.

My Products

Forge

Kompense

MarketingSoHigh

Believele

Let's Build
Together.

My Products

Forge

Kompense

MarketingSoHigh

Believele

Let's BuildTogether.

Let's Build
Together.