TESTS
Reproducible suites
Unit, integration, adversarial, and browser tests with recorded output.
Trust· claims matrix
Say what is true. Label the rest.
A trust center is only useful if it separates verified facts from targets, prototypes, and intent.
01 / POSTURE
Where the system actually is.
Version 1.0 is a specification. The private repository contains working prototypes with known gaps.
| Area | Current posture | Label |
|---|---|---|
| Bidding edge | Rust library and binary; local tests | PROTOTYPE |
| OpenRTB router | Rust service | PROTOTYPE |
| Control plane | Node HTTP and MCP | PROTOTYPE |
| MCP protocol | Handshake on older revision; 2026-07-28 is target | GAP |
| Provider access | No live production partner credentials observed | NOT CLAIMED |
| Certifications | None | NOT CLAIMED |
| Independent benchmarks | None | NOT CLAIMED |
02 / EVIDENCE
What a release must produce.
A claim is publishable when a reproducible test, a signed artifact, or an independent report backs it.
ARTIFACTS
Signed builds
Release metadata, checksums, and a verifiable supply chain.
REVIEW
Independent review
Security and financial-control review separate from the builder.
03 / SUBPROCESSORS
Published, not buried.
A current list of subprocessors, their purpose, location, and data categories belongs in the DPA and on this page.
The production subprocessor list is not published here because this is a design-stage site. The data processing terms define the obligation to publish it before onboarding.