Agent Governance Overview

Orlo governs agentic execution step by step without becoming an agent runtime.

What Orlo governs

• agent sessions
• governed step envelopes
• tool permission checks
• runtime limits
• approvals
• event timelines and live event streams
• trace samples and trace review
• trajectory results and step-level path metrics

What Orlo returns

At the API layer, Orlo can return:

• policy decisions for individual steps
• approval requirements and timeout behavior
• citations and retrieval evidence
• state change evidence
• session-level trajectory results
• step-level trajectory metrics

Why this matters

This lets teams do more than validate final text. They can review whether an agent:

• used the right tools
• respected approval boundaries
• changed state safely
• retrieved relevant context
• followed an acceptable path to the outcome

Scope boundaries

Orlo does not own:

• planning loops
• workflow orchestration
• long-lived agent memory systems
• general-purpose tool execution runtime

Adoption model

The most important developer surface is the SDK and adapter layer, but the machine interface remains the Platform API.