← All workLAB-003 / Working prototype / 2026
Pragma
The operations layer for Claude.
A local-first operations product that turns fragmented business signals into ranked, costed actions.
LAB-003Working prototype
Sector
AI operations
Services
Product strategy / Information design / MCP architecture / Dashboard UX / Prototype
Technology
Node.js / Model Context Protocol / Local dashboard / Data connectors
Stage
Working prototype with live and demo operating modes.
Give an AI assistant a disciplined operations layer: normalised metrics, explicit confidence and a prioritisation model grounded in goals and measurable value.
Pull read-only signals into a shared metric catalogue, score them against stage-aware bands and turn exceptions into actions ranked by impact, confidence and effort.
The dashboard makes health, goals, blind spots and next actions legible at a glance. The same state is available to Claude through MCP tools, so decisions can move between conversation and interface.
A zero-dependency Node.js CLI and MCP server coordinate connectors, local state and the dashboard. Live and demo inputs remain explicitly separated so plausible sample data can never masquerade as evidence.
PRODUCT DETAIL / SOURCE MATERIALWhat we explored
- AI-assisted operations
- Explainable prioritisation
- Local-first business intelligence
What we built
- CLI and MCP server
- Connector system
- Rules and scoring engine
- Interactive dashboard
Current stage
Working prototype with live and demo operating modes.
Next experiment
Expand connector coverage and validate the ranking model against more operating profiles.