We are a data company. We start by protecting yours.
Isolation is not a promise: it is a property of the infrastructure
The difference between “we filter by client in every query” and “the database does not allow anything else” is the difference between a convention someone can forget and a guarantee. We chose the latter at every point where it was possible.
One database per client, with its own role
Each client has its own database, its own role, and its own memory, with no permission to connect to anyone else's. Adding a client does not add shared infrastructure where its data could be crossed with others.
Views with the filter baked in
Read access goes through secure views that already carry the client filter inside them. Isolation is enforced by the database engine, not by the SQL of whoever queries.
Opaque identifiers
Bases, roles, and schemas use meaningless identifiers. Whoever holds a credential cannot enumerate who the other clients are: our portfolio cannot be inferred from the system.
Nothing listens on the internet unnecessarily
Data instances are not exposed to the internet. Read access goes through a proxy with mandatory TLS that only allows each client's read roles; the credential is issued once and rotated with a call.
Your source systems are read-only
Extract accesses your sources with read-only credentials, never writes to them, and does not leave files next to yours. What it brings in is deposited in a landing zone owned by your company, which the rest of the platform reads but does not modify.
Mandatory second factor and tamper-proof audit log
Two-factor access, strong password policy, brute-force throttling with alerts, and a hash-chained audit log where every decision is recorded.
AI is never the source of truth
Models write about facts that have already been calculated, and their output is validated: every table, number, and source must exist in the evidence. And external actions choose from a registered list, never from a free-form address.
Every data point knows where it comes from
Row-level lineage: which contract and which field produced each data point. Nothing is silently discarded, and every reprocess is idempotent.
About certifications
We prefer to be precise. iAgro does not hold its own information security certifications today. We do build on infrastructure and providers that maintain theirs, and the practices described above are implemented and verifiable in a technical assessment.
If your procurement process needs to review the architecture in detail, write to us: we will discuss it with your security team.
Does your procurement process need to review the architecture?
We can discuss it with your security team, in whatever detail you need.

