Authorize (Phase 2)

The Authorize phase defines what the agent is allowed to perform. Configure guardrails, policies, and behavioral rules to enforce governance.

Access via Agent Detail → Authorize tab.

Authorization Pipeline

Operations flow through three layers:

Incoming Operation
       │
       ▼
┌─────────────┐
│ Guardrails  │  Input/output validation and transformation
└─────────────┘
       │
       ▼
┌─────────────┐
│ OPA Policy  │  Stateless permission checks
└─────────────┘
       │
       ▼
┌─────────────┐
│ Behavioral  │  Stateful multi-step pattern detection
│ Rules       │
└─────────────┘
       │
       ▼
  Governance Decision

Sub-tabs

The Authorize tab has three sub-tabs:

Guardrails

Pre/post-processing validation and transformation:

Type	Purpose	Examples
Input Guardrails	Validate/transform incoming data	PII detection, rate limiting
Output Guardrails	Validate/transform responses	PII redaction, format enforcement

Creating a Guardrail:

1. Click Add Guardrail
2. Select type (Input or Output)
3. Configure trigger conditions
4. Define action (Block, Transform, Log)
5. Save and enable

Policies

OPA/Rego policies for stateless permission checks:

package openbox.agent.policy

default allow = false

allow {
    input.operation.type == "read"
    input.agent.trust_tier <= 2
}

allow {
    input.operation.type == "write"
    input.agent.trust_tier == 1
    input.context.approved == true
}

Creating a Policy:

1. Click Create Policy
2. Opens the Policy Editor
3. Write Rego code
4. Test against sample inputs
5. Save and attach to agent

Policy Context Available:

Field	Description
input.agent.id	Agent identifier
input.agent.trust_tier	Current trust tier (1-5)
input.agent.trust_score	Current trust score (0-100)
input.operation.type	Operation type (21 semantic types)
input.operation.target	Target resource
input.context	Session and request context

Behavioral Rules

Stateful rules that detect multi-step patterns:

Pattern	Example
Sequence	PII access → External API call (without approval)
Frequency	More than 10 failed auth attempts in 1 minute
Combination	Database write + File export + External send

Creating a Behavioral Rule:

1. Click Add Rule
2. Wizard opens with steps:
   • Trigger: What event starts tracking?
   • Conditions: What must happen?
   • Window: Time frame for pattern
   • Action: What happens on match?
3. Test with sample session data
4. Save and enable

Behavioral Rule Actions:

Action	Description
ALLOW	Permit and log
CONSTRAIN	Apply additional limits
REQUIRE_APPROVAL	Send to HITL queue
DENY_ACTION	Block the specific operation
TERMINATE_AGENT	Halt the entire agent session

Governance Decisions

The authorization pipeline produces one of five decisions:

Decision	Effect	Trust Impact
ALLOW	Operation proceeds	Positive (compliance)
CONSTRAIN	Proceeds with limits	Neutral
REQUIRE_APPROVAL	Pauses for HITL	Neutral (pending)
DENY_ACTION	Blocks operation	Negative
TERMINATE_AGENT	Halts session	Significant negative

Trust Tier-Based Defaults

Lower trust tiers receive stricter defaults:

Tier	Default Behavior
Tier 1	Most operations allowed, logging only
Tier 2	Standard policies enforced
Tier 3	Enhanced checks, some HITL
Tier 4	Strict controls, frequent HITL
Untrusted	All significant operations require approval

Next Phase

Once you've configured governance controls:

→ Monitor - Start your agent and observe its runtime behavior with Session Replay