Why each client has their own Ágora installation

There are two ways to serve software to companies. Multitenant SaaS, where thousands of clients share the same infrastructure and separation is by software. Or dedicated installation per client, where each client has their own database, engine and memory, separated at infrastructure level.

The mainstream industry chose multitenant a decade ago because it is more operationally efficient. The same devops team maintains a thousand clients. Economies of scale are brutal. For many use cases, multitenant is the right option.

Ágora goes the other way. Each client receives their own dedicated installation. It is not a minor decision, it is one of the design decisions that most impacts the whole operation.

Why we chose dedicated

Four reasons that reinforce each other. The first is data sovereignty. When a company entrusts us with sensitive information (commercial prospects, client history, audited critical decisions), that information should not coexist with another company's. Even less if that other company is a direct competitor, which in bounded markets like Argentina happens all the time.

The second is deep customization. Each client trains the agents with their own data, their own catalog, their own commercial rules, their own decision criteria. That specialization is not easily achieved in multitenant.

The third is operational independence. If a client wants to migrate away from OrvixLabs in the future, they take their complete installation, code, data, trained models. They are not tied to our infrastructure.

The fourth is resilience. If an installation has a problem, it does not affect the others. Each client is isolated from the failures of the others.

What it costs

This decision has real costs. Each deploy is more expensive. Each update requires coordination with each client, it is not done at once. Infrastructure costs are higher. Internal operation is more complex.

We take on those costs consciously because the client we target values them. A company that needs Ágora or IRIS is not the one that wants to pay ten dollars monthly per user. It is the one that wants real control over the artificial intelligence operating inside their company.

The commercial model

We charge fixed initial implementation according to complexity, plus monthly fee per system in operation. We do not charge per user or per usage volume. That means the client can grow without their monthly bill skyrocketing. The costs of commercial models (Claude, GPT, Gemini) are paid by the client directly to the provider, with their own API keys. That protects OrvixLabs from provider pricing oscillations and gives the client direct control over that expense.

For small cases this is more expensive than per-user SaaS. For serious cases, where information matters and operation is critical, it ends up much cheaper.