Agata Proxima Open-Source Rust Proxy Adds Tenant Isolation Layer for PostgreSQL

A developer has released the core engine of Agata Proxima, an open-source proxy written in Rust that sits between applications and PostgreSQL to enforce tenant isolation in multi-tenant deployments. The announcement was published on dev.to, where the author walks through the motivation, the architecture, and the decision to open-source the engine.

The problem being addressed

Multi-tenant SaaS products commonly store many customers' records in a single PostgreSQL database. Isolation between those customers is usually handled in the application layer, through authorization checks and tenant filters added to queries, and sometimes by PostgreSQL's Row-Level Security. As the dev.to post explains, each of these mechanisms can fail quietly: one missing filter, one incorrect permission check, or one misconfigured RLS policy can let one tenant read another tenant's data. The post's argument is that a second, independent enforcement point would catch exactly the class of mistake that application code is prone to.

Where the proxy fits

Agata Proxima inserts itself directly into the connection path: the application talks to the Rust proxy, and the proxy talks to PostgreSQL. Through that position, the engine enforces tenant-isolation rules and security policies, verifies access controls, and records database activity so administrators end up with auditable logs. The author is careful to frame the tool as an addition rather than a replacement: it is not meant to displace PostgreSQL's own security features, including Row-Level Security, but to add a defense-in-depth layer on top of them without changing the underlying database.

Why Rust

According to the post, Rust was selected for three reasons: memory-safety guarantees, strong performance, and its suitability for infrastructure software. The author's reasoning is that a component on the hot path between every application request and the database needs to be both fast and reliable, and Rust's compile-time guarantees are a natural fit for that role. The announcement includes no benchmarks, however, so any claims about performance overhead remain untested in public.

More than a proxy

The engine is only part of the picture. Around it, the developer has built a broader platform covering tenant management and policy configuration, access verification, database activity monitoring, audit records with what the post calls security attestations, engine node management, and developer APIs for integration. Notably, only the engine itself is explicitly described as open source; the post does not specify licensing for the surrounding platform. The engine source is published on GitHub in the Agata-Proxima organization, alongside a project website.

Questions the announcement leaves open

The author closes by asking PostgreSQL engineers what would concern them most about deploying such a proxy, naming performance overhead, connection management, compatibility, and operational complexity as candidate worries. Those are also the obvious evaluation criteria for any adopter: a proxy that terminates or mediates database connections changes failure modes, and every policy it enforces adds latency. With no published benchmarks, case studies, or adoption figures, the project reads as early-stage work actively seeking feedback from engineers who have handled tenant isolation in production.

Why it matters

Cross-tenant leakage is one of the most damaging failure modes in SaaS: it breaches every affected customer's trust at once, and it is typically caused by ordinary developer error rather than a sophisticated attack. A database-adjacent enforcement layer is a well-understood defense-in-depth pattern, and an open-source implementation in Rust makes that pattern inspectable and self-hostable rather than purely commercial. For teams running shared-database PostgreSQL architectures, a tool like this offers a way to catch a missing tenant filter before it becomes an incident, provided the performance and operational questions the author raises can be answered convincingly.