Getting Started

​

Open the Playground

• Left sidebar — Example policies and test scenarios
• Center — Monaco policy editor
• Right — Evaluation results

​

Write Your First Policy

1. Load a template — click “Templates” in the sidebar, pick one (e.g., “PII Detection”), click “Use Template”
2. Write from scratch — type a policy function in the editor

export default function myPolicy(request, context, response) {
  const piiResult = detectPII(response.content);

  if (piiResult.found) {
    return {
      action: 'DENY',
      reason: 'PII detected: ' + piiResult.types.join(', '),
      metadata: { piiTypes: piiResult.types }
    };
  }

  return {
    action: 'ALLOW',
    reason: 'No PII detected',
    metadata: {}
  };
}

​

Add Test Scenarios

• Name — descriptive label
• Prompt — the user input to test
• Provider — OpenAI, Anthropic, Gemini, or Bedrock
• Model — e.g., gpt-4
• Test type — pii, injection, or normal (determines mock response content)

​

Run Evaluation

• Decision — ALLOW, DENY, or MONITOR
• Execution time — how long the policy took
• Cost estimate — estimated LLM cost
• Token count — estimated tokens used

​

Export Your Policy

​

Available API Helpers

​

Keyboard Shortcuts

​

Self-Hosting

git clone https://github.com/agentguard-ai/TealTiger-Playground.git
cd TealTiger-Playground
npm install
npm run dev