Multi-tenant SaaS: how your data stays separate

Agentric is multi-tenant software: one platform, many businesses. That's what makes it affordable — and it only works if your data is truly sealed off from every other tenant. Here's how that works.

What "tenant" means here

Your organisation is a tenant: every record — calls, chats, contacts, bookings, knowledge — carries a tenant_id, and every query is scoped to it. There is no code path that returns data across tenants; the access layer itself enforces the boundary.

The isolation layers

• Database scoping — every model has tenant_id and every query filters on it. No exceptions, no shared tables for tenant data.
• Session binding — your login resolves to your tenant; requests for other tenants' resources return "not found" (not "forbidden" — we don't confirm they exist).
• Agent context — the assistant's knowledge, memory and tools are loaded per-tenant at call/chat start. A caller to tenant A's number can never reach tenant B's data.
• Widget scoping — the embed code carries your public tenant slug; it can only open conversations with your agent.
• Audit trail — administrative access (support, impersonation) is logged and visible.

What can't happen

• Another business's calls, customers or knowledge.
• Your conversation data feeding another tenant's agent.
• Shared credentials, shared memory, shared contact graphs.

What IS shared (deliberately)

Only the platform itself: the code, the AI provider connections, the infrastructure. Think of it as flats in a building — shared walls and plumbing, private rooms, your own keys.

Why this matters

Two reasons. First, privacy: your callers' data is yours, under your controller obligations. Second, trust: you can sign up to a shared platform without wondering who else can see your business. The isolation isn't a feature we describe — it's the load-bearing architecture.