Ops · Full-stack
PE Operations Control
The internal ops center for the Primo Engineering agent pipeline. Real-time visibility across tickets, QA, AI cost, and agent jobs. Built it because running a team of AI agents without it is not manageable.
Stack React · Express · SQLite · Docker · Claude
· Internal tool
The problem
Running a team of AI agents across several repos generates work faster than any one person can track: tickets, QA results, AI spend, and dozens of agent jobs at once. Without a single place to see it, the operation stops being manageable.
What it does
PE Operations Control is the internal dashboard for that pipeline. It gives real-time visibility across tickets, QA, AI usage and cost, and every agent job, and it dispatches agents as sandboxed containers so the whole operation runs from one screen.
How it works
- A React, Express, and SQLite app in Docker on a VPS, dispatching agents as isolated containers.
- AI ticket generation with Claude that returns structured output with compliance flags and effort estimates.
- A QA tracker with CSV import, filtering, and one-click YouTrack bug creation.
- AI usage and cost monitoring across Anthropic and OpenAI, with budget alerts.
- Session auth with AES-256-GCM encrypted credentials, so API keys never reach the browser.
The value
- One pane of glass for an entire agent operation.
- Cost stays under control with live spend tracking and budget alerts.
- Credentials are encrypted and never exposed to the client.
- Built and used in production daily, not a demo.