Skip to main content
Platform architecture

Enterprise trust. One intelligence layer.

CataZenta is built for enterprise trustAPI-first, multi-tenant, and designed to feed classical integrations and AI agents from the same governed catalog.

The web UI, syndication workers, specialist agents, Zen AI, and MCP all call the same REST surface — no shadow APIs, no separate chatbot database.

API-firstUI · workers · MCP share /v1
Multi-tenantOrg → tenant · RBAC
AgenticACC · Zen AI · MCP

Platform at a glance

LayerResponsibility
ExperiencePIM UI, Zen AI chat, Agent Control Center
IntegrationREST /v1/*, MCP tools, webhooks, async jobs
CatalogProducts, attributes, categories, DAM, completeness
SyndicationConnections, export profiles, per-SKU job items
AIPlatform-managed models (Zen), specialist agents, MCP tools on REST
GovernanceRBAC, workflows, approvals, audit

Who we serve → Commerce ecosystem


Architecture principles

Eight docs — each answers a question enterprise architects and integrators ask.

Multi-tenant

Organization → tenant isolation with RBAC and scoped JWT claims.

AI infrastructure

Tenant-scoped models, audit, human review — not a public chatbot on CSV exports.

Scalability

Async import/export jobs at catalog scale; workers, not synchronous bulk HTTP.

Security & trust

Data protection, isolation, AI governance — materials for enterprise review.

Security architecture

Auth model summary for technical reviewers and procurement.

Data model

How catalog, media, connections, and export jobs connect.

Zen AI & MCP

Zen AI, MCP, and specialist agents on one catalog — identity and isolation.


Agentic PIM in the architecture

Specialist agents are first-class platform citizens — not a bolt-on chat widget.

ComponentRole
Agent Control CenterEnable agents, triage insights, approve writes
Zen AIConversational assistant in PIM on governed catalog data
MCP endpointStructured tools for external AI clients — same REST surface
SyndicationConnections, export profiles, per-SKU job results

Agentic PIM · Agent Control Center


Three AI paths, one catalog

PathEntryIdentityModel
Zen AIChat in PIMPIM sessionPlatform-managed when using CataZenta AI credits
Hosted MCPhttps://api.catazenta.com/mcpOAuth JWTClient model (Claude, GPT)
Specialist agentsAgent Control Centerai_mgmt + tenant policyPer tenant agent policy

All AI paths execute catalog operations with the caller's permissions on the same REST APIs.

MCP overview · Zen AI


Data & syndication

EntityArchitecture role
Product familySchema for validation, AI extraction, channel mapping
CompletenessPublish gates and agent thresholds
ConnectionChannel credentials (Amazon LWA, Shopify, …)
Export job + itemsPer-SKU syndication truth and retry scope

Expanded ER diagram → Data model · Setup concepts → Catalog fundamentals

Syndication flow → Integrations · Syndication guide


Start by role


Platform context

Traditional PIM concepts still exist under the hood — they power intelligence rather than defining the story:


Get started

  1. Read API-first — confirm no shadow APIs for your integration design.
  2. Review Multi-tenant — tenant scope for sandbox vs production.
  3. Walk Zen AI & MCP if AI is in scope.
  4. Share Security & trust with procurement.
  5. Map entities via Data model before custom connectors.