Rules

Rules are the governance layer of LensOS. They let you define behavioral boundaries, content policies, and response guidelines for your agents — without modifying application code.

How Rules Work

Rules are evaluated at runtime before and after each agent response. They can filter inputs, constrain outputs, and enforce compliance policies. Rules are applied in priority order, and you can combine multiple rules to create layered policies.

Rule Types

Content Rules

Control what topics and content your agent can discuss:

{
  "type": "content",
  "action": "block",
  "description": "Block discussions about competitors",
  "patterns": ["competitor-name", "alternative-product"]
}

Behavior Rules

Define how your agent should behave and respond:

• Response tone and formality level
• Language and localization preferences
• Citation and source requirements
• Response length constraints

Safety Rules

Built-in safety rules protect against common issues:

• PII detection and redaction
• Harmful content filtering
• Prompt injection protection
• Rate limiting per user

Managing Rules

Rules can be managed through the dashboard or the API:

await lens.rules.create({
  name: 'concise-responses',
  type: 'behavior',
  description: 'Keep responses under 200 words',
  priority: 10,
  active: true,
})