Audit before assertion.
Every agent shows its work — file citations, token caps, append-only audit logs. If we cannot trace it, we will not ship it.
About
Building the AI cortex of every company.
Cortex is the flagship product of Pleisys, the AI subsidiary of Verdeshell Technologies Pvt Ltd. We exist for one reason: every founder and operator deserves a full C-suite of grounded advisors, before they can afford to hire one.
The story
Verdeshell Technologies Pvt Ltd runs a portfolio of SaaS products and a services engagement. Across that portfolio, the same pattern kept repeating: a founder or operator made a cross-functional decision with one perspective when they needed seven.
We started building an internal markdown-based agent system to pressure-test our own decisions — a CEO, CFO, CTO, CISO, CMO, COO, CDO grounded in the company’s real context files. It worked. It scaled. It compounded.
Cortex is the productised version. Same architecture: read the context, run the CxO seat, cite the files, write a memo. Built multi-tenant, audit-traceable, and compliance-ready from day one — because the moment another founder uses it, our standards have to be theirs.
Brand hierarchy
Parent company
Verdeshell Technologies Pvt Ltd
Operates the Verdeshell ERP suite (CRM OS, Career, Recruit, Courier Hub, HRMS) and a services arm.
AI subsidiary
Pleisys
Houses the Pleisys AI suite: Echo-IQ (speech), VerdePulse (engagement), GenVision (vision), chat.pleisys.com (conversational), and Cortex — the flagship.
Flagship product
Cortex
The complete AI-driven C-suite. The AI cortex of every company.
What we believe
Grounded over generic.
A C-suite agent that does not read your context is roleplay. We make the topic-file library load-bearing, never optional.
India-built, world-shipped.
Razorpay + Stripe day one. DPDP + GDPR + CCPA legal pack day one. Single-tenant per-row isolation day one. No "Phase 2" excuses.
Compounding decisions.
Every memo lands in the library. Future synthesis reads the past. Cortex gets smarter at your business with use, not at LLM-generic vibes.
The build
Twelve sprints from a markdown agent system to a production-ready, audit-traceable SaaS.
- Sprint 0Vendor + infra picked. Anthropic Sonnet 4.6 + Postgres 17 + Next.js 16 locked.
- Sprint 1Tenant isolation foundation — JWT + ALS tenant context + FORCED Postgres RLS.
- Sprint 2Authentication, signup, refresh-token rotation.
- Sprint 3Context library — agent roles, products, topic files.
- Sprint 4Single-agent chat with Prompt Caching + SSE.
- Sprint 5Multi-agent synthesis + decision memos.
- Sprint 6Quotas + subscription lifecycle (trial → grace → suspended).
- Sprint 7Payments — Razorpay (INR + GST) + Stripe (USD zero-rated).
- Sprint 8Legal pack, DSAR self-service, account closure, accessibility.
- Sprint 9VCP operator console, impersonation audit trail, Web Vitals, k6 load test. v1.0-RC.
- Sprint 10Marketing site, public launch prep.
Want the technical detail?
Our security page covers tenant isolation, encryption, audit logging, and sub-processor disclosure. Our legal pack covers DPDP + GDPR + DPA + cookies.