NapsixAI
Your own AI
Case study · B2B marketplace

NapsixB2B: we use it first ourselves

Our own B2B marketplace: it connects SMBs with business buyers, with AI agents wired to the catalog, stock and business data. Napsix.com reached 210,000 users.

+20
MCP servers
9
SMB sectors
3
Subdomains
16.2
Next.js
NapsixB2B

What it is NapsixB2B

NapsixB2B is the industrial marketplace we use first ourselves: SMBs selling to business buyers, with AI agents that query the catalog, stock and business data.

It is also our test bench: every pattern we apply for clients was proven here first.

The challenge

  • A B2B marketplace that needed agents connected to real data, not a generic chatbot.
  • Multiple products (marketplace, admin panel, agentic chat) with a single session.
  • Live integrations with monetary and financial system data.
  • Security and plan gating without friction for the user.

The NapsixAI solution

  • A conversational assistant with +20 integrated MCP servers.
  • Three subdomains sharing one session: marketplace, panel and agentic chat.
  • Live integrations with monetary, exchange-rate and debtor data.
  • Starter, Pro and Enterprise plans with feature flags per organization.

Technology stack

How it is built.

Next.js 16.2
App Router + RSC + Turbopack
Neon Auth + Neon Postgres
Per-feature branching + RLS
MCP
+20 servers integrated in the assistant
Drizzle ORM
Multi-tenant + audit logs
Tailwind v4 + shadcn/ui
Consistent design system
Vercel AI Gateway
Multi-model (OpenAI, Anthropic)

What we deliver

What makes the difference day to day.

AI assistant per SMB

A conversational assistant with +20 MCP servers. Each SMB has its agent with access to tools, catalog and panel.

Live integrations

Monetary and exchange-rate data, rejected checks and debtors from the Argentine financial system, in real time.

Multi-tenant + subdomains

Marketplace, panel and agentic chat with the same Neon Auth session.

Security and gating

JWT iframe forwarding, CSP frame-ancestors and plans with feature flags per organization.

Business data wired in

The agent queries real stock and catalog before answering.

A base for clients

The same system runs client brands with their catalog and their team.

Do you have a similar project?

Tell us what your operation needs and we will look at what to build first.