CAGE

CAGE provides structure for agentic software engineering. It’s part of Business Engineering and AI-Native organizational design.

The problem with agents

Frontier models and harnesses are advancing at breakneck speeds. It’s exhausting to keep up.

Some of the current problems with agents are:

• Agent capability does not provide accountability
• Sessions do not provide durable memory
• Confidence does not provide authority
• Automation does not decide where human judgment is required

When applying AI to the AI-Native organization, these problems are multiplied.

The CAGE model

CAGE is built to help your business focus on outcomes. It provides operating structure for humans and agents working together.

• Defined agent roles and authority boundaries
• Durable artifacts stored outside individual sessions
• Directives that define session scope
• Explicit build modes: autonomous, supervised, and guided
• Governed workflows for implementation and validation
• Human adjudication of changes to governing instructions
• Provider-neutral contracts and replaceable tools