Skip to main content
Test policies before deployment to ensure they work as expected. Simulation is mode-specific: the Simulate tab uses the same mode as the policy (Conditions, Instructions, or Guardrails), and the payload shape and evaluation result format depend on that mode.

​
Testing Policies with Sample Payloads

  1. Open a policy in the editor
  2. Navigate to the “Simulate” tab in the right sidebar
  3. Enter a sample payload (format depends on policy mode — see below)
  4. Click “Evaluate” to see how the policy processes the payload

​
Mode-specific payloads

The Simulate tab sends the payload to the backend based on the policy type. Use the shape that matches your mode:
ModePayload shapeExample
ConditionsStructured object (e.g. request, user){ "request": { "amount": 5000, "currency": "USD" }, "user": { "id": "user-123", "risk_level": "high" } }
InstructionsPrompt plus response (or input){ "prompt": "User question...", "input": "LLM response text..." } — the field names may match the Instructions evaluate API.
GuardrailsText to checkA string or object with a single text field (e.g. the model output or user message to scan).
The evaluation trace and result (decision, reason, matched rules) may vary by mode; for example, Conditions show structured condition evaluation, while Guardrails show which guardrail type triggered.

​
Quick Test Templates

The simulator provides quick templates for common scenarios:
  • Basic Request: Simple request with amount and user
  • High-Value Transaction: Large amount transaction
  • High-Risk User: Request from high-risk user profile
  • Custom: Create your own payload

​
Custom Payload Input

Enter a payload that matches your policy mode. For Conditions, a typical payload is a JSON object with request and optional user (or other fields your conditions reference): 
{
  "request": {
    "amount": 5000,
    "currency": "USD"
  },
  "user": {
    "id": "user-123",
    "risk_level": "high",
    "country": "US"
  }
}
For Instructions or Guardrails, use the prompt/input or text shape described in the table above.

​
Evaluation Trace Visualization

After evaluation, you’ll see:
  • Decision: Allow, Block, or Escalate
  • Matched Conditions: Which conditions evaluated to true
  • Evaluation Steps: Step-by-step trace of the evaluation
  • Execution Time: How long the evaluation took

​
Simulation History

  • Previous simulations are saved in the history
  • Re-run previous simulations
  • Compare results across different payloads

​
Unit Tests Display

View and manage unit tests for the policy:
  • Predefined test cases
  • Expected outcomes
  • Run all tests to validate policy behavior

​
Real-time Condition Evaluation

As you modify conditions, you can:
  • Test immediately with the current payload
  • See how changes affect the evaluation
  • Iterate quickly to refine your policy

Policies

Learn about policy management

Policy Conditions

Understand condition logic