An agent-driven incident response system: it investigates operational alerts, gathers evidence, proposes remediation, requests human approval for risky actions, executes approved runbooks, and verifies recovery.

An enterprise gateway that safely exposes internal capabilities to AI agents through MCP — every tool call passes through authentication, policy-based authorization, schema validation, risk classification, and audit logging.

A framework for repeatedly testing agent behavior under normal and adversarial conditions — it answers the question every agentic system dodges: how do we know this agent is actually reliable?

A portable desktop PR dashboard for teams that work in sprints across many repos in one GitHub organization — one window that answers: what’s open, what needs review, what has changes requested, what’s approved, and what merged this sprint.