Opus Control

Opus Control is a real-time, AI-native mission-control dashboard that monitors an EC2 instance via Redis, uses the Anthropic (Claude) API to analyze load and suggest resource allocation, and lets you control resources by dragging an allocation line on a live graph (up = more resources, down = less resources). The interface breaks usual patterns by making the chart the main control (drag-to-allocate, with Claude able to adjust it), showing anomalies in a modal with editable reasoning and throttle sliders rather than a chat thread, and using a collapsible Context panel to shape what Claude sees (watch/ignore, thresholds, time window). The innovation is that Claude’s reasoning drives the state of the UI anomaly alerts, rephraseable explanations, and the moving allocation line, so the dashboard itself is the shared reasoning and control surface, with user overrides and dismissals stored and fed back. Built with Python, FastAPI, uvicorn, Redis, Anthropic SDK, and matplotlib on the backend; React, Vite, TypeScript, Tailwind CSS, and Recharts on the frontend. These choices give one server for REST and WebSockets, live graph and allocation updates, and async Claude calls so the UI stays responsive while the model drives what appears and what you can do, human AI interaction as a shared control loop, not a single chat window.