Series · 4 posts · about 40 min
AI gateway
How to govern model and agent traffic through a gateway: operating models, routing, a cross-cloud containment lab, and where policy should bind.
-
Build the Gateway, Buy the Model: AI Governance for Regulated Industries on GCP
Self-hosting an open model is not what keeps your data safe in a regulated industry - owning the front door is. I built an AI governance gateway on GCP with PII screening, token budgets, and a full audit trail for under 10 dollars a month, and it changed my answer on the whole build-vs-buy question.
-
Stacking Governance Layers Without Losing Track of Them
Seven or eight layers all answering some version of "is this allowed". Two questions sort almost all of it: where does each one bind, and who can actually change it.
-
Choosing an AI gateway operating model
The AI gateway decision is about how many data planes sit under one control plane, and model traffic and agent tool traffic answer it differently. This sets out the three operating models, a routing rule for placing traffic, and the standard a local gateway must meet to carry it.
-
One registry, one gateway, two clouds: testing AI agent containment across AWS and Google Cloud
A lab across AWS and Google Cloud in which agents on both clouds send every model call and MCP tool call through one gateway that checks one registry. Disabling one record stopped that agent's traffic on either cloud; the post gives the design, code, results, latency and data transfer measurements, and what was not tested.