Skip to content

QUICKSTART.md

Latest commit

 

History

History
113 lines (89 loc) · 2.95 KB

QUICKSTART.md

File metadata and controls

113 lines (89 loc) · 2.95 KB

AgentGuard Quick Start 🚀

Get protection in 30 seconds.

1. Install

# NPM (recommended)
npm install agent-guard

# Or download directly
curl -O https://raw.githubusercontent.com/dipampaul17/AgentGuard/main/agent-guard.js

2. Add to your agent

const agentGuard = require('agent-guard');
await agentGuard.init({ limit: 10 });

// Your existing agent code...
// AgentGuard now protects everything below

3. Verify it works

node verify-installation.js

Watch the cost counter: 💸 $0.12... $0.45... $0.89...

4. Configure for your needs

const guard = await require('agent-guard').init({
  limit: 50,                                    // Kill at $50
  webhook: 'https://hooks.slack.com/...',      // Slack alerts
  silent: false                                // Show cost updates
});

// Check costs anytime
console.log(`Current: $${guard.getCost()}`);

5. For browser agents

<script src="agent-guard.js"></script>
<script>
  AgentGuard.init({ limit: 25 });
  // Your browser agent code...
</script>

Real Example: RAG Agent Protection

// Add at the top of your RAG agent
await require('agent-guard').init({ limit: 100 });

// Your existing code
async function processDocument(doc) {
  const chunks = await chunkDocument(doc);
  
  for (const chunk of chunks) {
    // These API calls are now protected
    const embedding = await openai.embeddings.create({...});
    const response = await openai.chat.completions.create({...});
    
    // AgentGuard tracks each call automatically
    console.log('Processed chunk:', response);
  }
}

Emergency Stop Examples

Runaway Loop Protection

💸 $0.23... $1.45... $4.67... $9.89
⚠️  AGENTGUARD: Approaching limit ($10)
🛑 AGENTGUARD: KILLED PROCESS - Saved you $90

What Gets Tracked

  • ✅ OpenAI API calls (all models)
  • ✅ Anthropic/Claude API calls
  • ✅ Any API response logged to console
  • ✅ Fetch/axios requests to AI APIs
  • ✅ Real-time token & cost calculation

What Gets Protected

  • 🛡️ Infinite loops calling AI APIs
  • 🛡️ Expensive model calls (GPT-4, Claude Opus)
  • 🛡️ Accidentally high token usage
  • 🛡️ Recursive agent calls
  • 🛡️ Development debugging sessions

Troubleshooting

"Not tracking my API calls"

  • Make sure you console.log() the API responses
  • Or use fetch/axios for automatic interception

"Kills too early"

  • Adjust limit: guard.setLimit(100)
  • Check pricing accuracy for your model

"Want to exclude certain calls"

  • Initialize multiple guards for different components
  • Use silent: true mode for non-critical calls

Next Steps

Stop losing money. Start shipping safely. 🛡️