Cortex Console

PROBLEM

Autonomous agents can write code and execute tools, but they often operate with weak role separation, insufficient review, and overly broad permissions. That makes it difficult to trust what will be executed before changes reach the project.

WHAT I BUILT

A local desktop console that separates the workflow into specialized agents: Explorer explores solutions, Critic challenges proposals, Architect turns approved decisions into architecture and tasks, Builder implements only authorized work, and Verifier reviews the result and runs deterministic checks. Local execution uses separated models through Ollama.

RESULT

An AI-assisted development environment where exploration, decision-making, implementation, and verification no longer happen inside a single unrestricted agent. Cortex Console also acts as the control surface for the broader CortexOS architecture focused on agent policies, permissions, isolation, and auditability.

WHAT IT DOES

• Multi-agent workflow with separated responsibilities
• Explorer, Critic, Architect, Builder and Verifier
• Local model execution through Ollama
• Human approval before implementation
• Deterministic verification before merge
• Desktop application built with Tauri
• Local-first architecture
• Foundation for an agent policy and security runtime

BUILT WITH