Profiles Overview

Profiles define how guardrails are composed, configured, and executed together.

A profile represents a policy bundle that controls:

  • Which guardrails run
  • In what phase (input, output, tool, operational)
  • With what configuration
  • For which use case

Profiles are the primary abstraction used by Guardrails to apply safety, security, and compliance rules consistently across requests.

Why profiles exist

Without profiles, guardrails would need to be:

  • Manually enabled per request
  • Reconfigured repeatedly
  • Hard to reason about at scale

Profiles solve this by providing:

  • Reusable configurations
  • Centralized policy management
  • Clear separation of concerns
  • Production-ready defaults

What a profile contains

At a high level, a profile includes:

  • Metadata (name, description)
  • Input guardrails
  • Output guardrails
  • Tool guardrails
  • Optional operational guardrails

Each guardrail is defined declaratively with optional configuration.

When profiles are applied

Profiles are resolved and compiled:

  • At request time
  • Based on profileName or profileId
  • Before guardrail execution begins

The compiled profile is then executed by the Guardrails engine.

Profile lifecycle

  1. Profile is defined (built-in or custom)
  2. Profile is selected at runtime
  3. Profile is compiled into executable guardrails
  4. Guardrails are executed
  5. Analytics are recorded

Next steps

  • Learn about built-in profiles → Built-in Profiles
  • Create your own policies → Custom Profiles
  • Understand compilation → Profile Compilation