Documentation

Quick Start

Get started with AI Control Tower in less than 10 minutes. Track your first AI decision and start building compliance from day one.

Installation

JavaScript/TypeScript

npm install @ai-control-tower/sdk

# or

yarn add @ai-control-tower/sdk

Python

pip install ai-control-tower

Other Languages

Use our REST API directly. SDKs for Go, Ruby, and Java coming soon.

Authentication

Initialize the SDK with your API key:

JavaScript/TypeScript

import { AIControlTower } from '@ai-control-tower/sdk';

const act = new AIControlTower({
  apiKey: process.env.ACT_API_KEY,
});

Python

from ai_control_tower import AIControlTower

act = AIControlTower(api_key=os.environ["ACT_API_KEY"])

Integrity Verification

AI Control Tower uses SHA-256 hash-chain technology to create tamper-evident audit trails for all AI decisions. Every decision is cryptographically linked to the previous one, making it impossible to modify historical records without detection.

Verify Chain Integrity

const verification = await act.verifyIntegrity({
  startDate: "2024-01-01",
  endDate: "2024-01-31",
});

console.log(verification.status); // "valid" | "compromised"
console.log(verification.decisionsVerified); // 1,247
console.log(verification.chainHash); // "sha256:abc123..."

Get Decision Hash

const decision = await act.getDecision("dec_abc123");

console.log(decision.integrity);
// {
//   hash: "sha256:7f83b162...",
//   previousHash: "sha256:a4bfc3d2...",
//   chainPosition: 1247,
//   verified: true
// }

Usage Monitoring

Track AI usage across your organization with governance-focused monitoring. Set policies, monitor compliance, and receive alerts when thresholds are exceeded.

Get Usage Statistics

const usage = await act.getUsage({
  period: "30d",
  groupBy: "user", // or "model", "department"
});

console.log(usage.totalDecisions); // 12,847
console.log(usage.activeUsers); // 45
console.log(usage.policyViolations); // 0
console.log(usage.breakdownByUser); // [...]

Set Usage Alerts

await act.createAlert({
  type: "usage_threshold",
  metric: "decisions_per_user",
  threshold: 1000,
  period: "daily",
  severity: "warning",
});

// Also available:
// - "cost_threshold" - Alert when costs exceed limit
// - "high_risk_decision" - Alert on high-risk AI decisions
// - "integrity_violation" - Alert if chain verification fails

Track AI Decision

Track every AI decision with full context, inputs, outputs, and metadata:

const decision = await act.trackDecision({
  model: "gpt-4",
  provider: "openai",
  input: "What is the capital of France?",
  output: "The capital of France is Paris.",
  metadata: {
    userId: "user_123",
    sessionId: "session_456",
    environment: "production",
  },
  cost: 0.0023,
  latency: 234,
  tokens: {
    input: 8,
    output: 7,
    total: 15,
  },
});

console.log("Decision ID:", decision.id);

Generate Explanation

Generate human-readable explanations for AI decisions to meet regulatory requirements:

const explanation = await act.explainDecision({
  decisionId: decision.id,
  framework: "gdpr", // or "hipaa", "eu-ai-act", etc.
});

console.log(explanation.summary);
// "This AI decision used GPT-4 to answer a factual question..."

Check Policy Compliance

Validate decisions against your governance policies before execution:

const policyCheck = await act.checkPolicy({
  model: "gpt-4",
  provider: "openai",
  input: userPrompt,
});

if (policyCheck.status === "violation") {
  console.error("Policy violation:", policyCheck.violations);
  // Block the AI request
} else {
  // Proceed with AI request
  const response = await openai.chat.completions.create({...});

  // Track the decision
  await act.trackDecision({...});
}

REST API

Use the REST API directly if you prefer:

Base URL

https://api.aicontroltower.com/v1

POST /decisions

Track a new AI decision

curl -X POST https://api.aicontroltower.com/v1/decisions \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4",
    "provider": "openai",
    "input": "What is the capital of France?",
    "output": "The capital of France is Paris.",
    "cost": 0.0023,
    "latency": 234
  }'

POST /explanations

Generate an explanation for a decision

curl -X POST https://api.aicontroltower.com/v1/explanations \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "decision_id": "dec_abc123",
    "framework": "gdpr"
  }'

POST /policy-checks

Check if a decision would violate policies

curl -X POST https://api.aicontroltower.com/v1/policy-checks \
  -H "Authorization: Bearer YOUR_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "gpt-4",
    "provider": "openai",
    "input": "Sensitive user prompt"
  }'

Webhooks

Configure webhooks to receive real-time notifications for policy violations and other events:

Event Types

• policy.violation - A decision violated a policy
• cost.threshold - Cost threshold exceeded
• decision.created - New decision tracked

Webhook Payload

{
  "event": "policy.violation",
  "timestamp": "2024-01-20T15:30:00Z",
  "data": {
    "decision_id": "dec_abc123",
    "policy_id": "pol_xyz789",
    "severity": "high",
    "violation_details": {
      "rule": "cost_limit",
      "threshold": 0.01,
      "actual": 0.023
    }
  }
}

Configure webhooks in your notification settings.