Agent Governance Console
One place to see what the agents did, what they decided, and whether the practice behind these builds is healthy
Two-minute walkthrough of the console
Cost & Performance — real spend on my own account, tracked per session and per agent role
RAG Analytics — index drift detection across the ecosystem's knowledge bases
The Intelligence view — proposal throughput, conversion, and the stale-proposal count the pipeline is judged on
Running a multi-agent stack across several projects produced governance data of every kind: audit logs, impact classifications, hook events. All of it sat in flat files scattered across different repositories, with no single place to ask whether the ecosystem was healthy or what the agents had actually done that day.
Surveyed several production observability and evaluation tools (LangSmith, Braintrust, Arize Phoenix, and others) to understand what they did well and where they were over-built for a personal use case.
A Next.js dashboard backed by PostgreSQL, ingesting governance data from every project spoke and wrapped in an Electron menubar shell. An append-only audit log, decision tracking with reasoning capture, and multi-factor health scoring. It runs and records today; active development stopped once it had taught me what I built it to learn.
- An append-only audit trail, decision tracking with reasoning capture, and multi-factor health scoring across every project
- Health scoring that hard-caps any section lacking independent evidence, so the number cannot drift upward on assertion
- Scoped to a few views backed by real data, each one answering a question I actually asked during a session