For Engineering

GenZAgents for enterprise architects — the AI metalayer in your architecture diagram

Your AI strategy assumes a vendor-neutral metalayer to avoid lock-in and enable cross-provider portability. GenZAgents is that metalayer — a documented architectural slot in your AI reference design.

The metalayer slot

Reference AI architecture: (1) Model providers — Anthropic, OpenAI, Google, Azure, self-hosted. (2) Agent frameworks — LangChain, CrewAI, AutoGen, custom. (3) IDE / chat clients — Claude Desktop, Cursor, ChatGPT, in-house apps. (4) The metalayer above all of them — receipt issuance, audit trail, cross-provider portability, governance. GenZAgents fills slot (4).

Why this slot can't live inside a provider

Anthropic / OpenAI / Google profit from continuity inside their own stack. Cross-provider portability damages their lock-in story. They won't build it; same logic as Plaid not being built by JPMorgan. The metalayer has to be vendor-neutral by structural necessity, and the buyer side is what funds it. That's GenZAgents' market position.

How GenZAgents integrates with your architecture

Receipt issuance: MCP server (IDE-side) + SDK (in-house apps). Audit trail: dashboard + evidence packs + webhooks → SIEM. Cross-provider portability: portable manifest + restore_chat. Governance: per-project ACLs + anomaly detection + dispute resolution. Each integration point is documented and stable; the metalayer is a buildable architectural slot.

Where it sits in your reference architecture diagram

Bottom layer: model providers. Middle layer: agent frameworks. Top-of-stack: IDE / chat / in-house apps. Side-channel metalayer (drawn off to the right): GenZAgents — receives signed events from all layers, emits receipts, surfaces audit + governance. Standard pattern; cleanly drawable; explains the architecture in one slide.

Self-hosted vs SaaS architectural decision

SaaS: simpler ops, faster updates, multi-tenant cost share. Self-hosted: data residency, air-gap compatibility, predictable cost. Pick based on your data-classification policy + regulatory exposure. We support both; the architectural slot is the same.

Integration with your existing identity provider

SSO via Microsoft Entra or Google Workspace (standard SAML / OIDC). Domain-verified auto-membership for engineers. Per-project ACLs map to Entra groups. SCIM is on the Enterprise tier (auto-provisioning from your IdP).

Common questions

How does this fit with our existing API gateway / Kong / Apigee?

Independent. GenZAgents sits at the agent / IDE side, not at the API edge. The integration is via SDK from your in-house apps, not via API gateway interception.

Can I run this in a multi-cloud setup?

Yes — the SaaS service is on Azure UK South; the Helm-chart-based self-hosted option works on any K8s (AWS EKS, GCP GKE, Azure AKS, on-prem). Mix and match as architecture requires.

Where does the long-term receipt storage live?

SaaS: Azure Postgres + Supabase. Self-hosted: your Postgres. The receipt schema is open (work-receipt-spec on GitHub); you could even archive to your own data lake while we serve the operational queries.

How does this integrate with our existing data governance?

Per-receipt project tags map to your data-governance categories. The receipt feed feeds into your data-governance dashboards (Collibra, Alation, etc.) via CSV export or API pull.

Related

Get the trust layer for your AI work

GenZAgents is the verified work-history layer above every AI provider your team uses. Sign cryptographic receipts, hand off conversations across Claude / ChatGPT / Cursor / Gemini, keep institutional AI knowledge when employees leave.

Last reviewed · 2 min read· Open spec· Changelog