Edgecase Labs — ExceptionOS

Every edge case becomes automation.

ExceptionOS is an AI-powered operational exception-handling system for e-commerce teams. It turns a human correction for a new edge case into reusable procedural knowledge, so analogous future cases can resolve autonomously with an auditable decision trail.

Architecture

Next.js + TypeScript powers the dashboard and API surface; LangGraph controls the deterministic case workflow. Band provides multi-agent collaboration, Terac MCP handles human escalation, Linq provides messaging intake and updates, and Render Workflows runs case execution.

Three specialists

• Memory Specialist retrieves policies and learned skills.
• Resolution Specialist proposes an operational action from the available evidence.
• Policy Critic independently approves or blocks the proposed action.

Golden demo

Unknown exception → human correction → learned skill → analogous case resolves autonomously

The deterministic dashboard demo shows the critic escalating an unsupported case, compiles the correction into a skill, retries the original case, then avoids human intervention on a distinct analogous case.

Run locally

npm install
npm run dev

npm test
npm run typecheck
npm run build

Live app

Render URL: TBD

Hackathon sponsor integrations

• Band — remote multi-agent collaboration
• Terac MCP — human-in-the-loop escalation
• Linq — SMS/iMessage/RCS operational messaging
• Render Workflows — durable workflow execution