Costs & Credits

Fresh Source: docs.airtop.ai

Airtop uses a credit-based billing system. Credits cover AI processing, CPU time, and proxy bandwidth — not tokens.

What Costs Credits

Resource	Cost Driver
Sessions	Billed per 30-second increment of running time
AI operations	pageQuery, scrapeContent, fillForm, monitor, etc.
Proxy bandwidth	Data transferred through residential proxies
Captcha solving	Built-in solver attempts

Cost Controls

Set limits to prevent runaway spending on individual operations or sessions.

Per-Operation Limits

TypeScript

const result = await client.windows.pageQuery(sessionId, windowId, {
  prompt: 'Extract all product data',
  configuration: {
    costThresholdCredits: 10,    // Max 10 credits for this operation
    timeThresholdSeconds: 30,    // Max 30 seconds for this operation
  },
});

Python

result = client.windows.page_query(
    session_id, window_id,
    prompt="Extract all product data",
    configuration={
        "costThresholdCredits": 10,
        "timeThresholdSeconds": 30,
    }
)

Per-Session Limits

TypeScript

const session = await client.sessions.create({
  configuration: {
    timeoutMinutes: 10,  // Session auto-terminates after 10 minutes
  },
});

Python

session = client.sessions.create(
    configuration={"timeoutMinutes": 10}
)

Cost Optimization Tips

1. Terminate Sessions Promptly

Sessions bill per 30-second increment. Always terminate when done.

try {
  // ... your operations
} finally {
  await client.sessions.terminate(sessionId);
}

2. Reuse Sessions for Multiple Operations

Opening a new session for each URL is wasteful. Use multiple windows in one session.

// Efficient: one session, multiple windows
const session = await client.sessions.create({ configuration: { timeoutMinutes: 15 } });
const sessionId = session.data.id;

for (const url of urls) {
  const window = await client.windows.create(sessionId, { url });
  await client.windows.scrapeContent(sessionId, window.data.windowId);
  await client.windows.close(sessionId, window.data.windowId);
}

await client.sessions.terminate(sessionId);

3. Use scrapeContent Instead of pageQuery When Possible

scrapeContent is cheaper than pageQuery because it doesn't require LLM inference for the extraction itself.

4. Set Cost Thresholds

Always set costThresholdCredits and timeThresholdSeconds in production to prevent unexpected costs.

5. Batch Operations Efficiently

Use appropriate maxConcurrentSessions — more parallelism = faster but more concurrent session billing.

Monitoring Usage

Check your credit usage in the Airtop Dashboard:

• Usage tab — current period credit consumption
• Session history — per-session cost breakdown
• API logs — per-operation credit usage

Best Practices

1. Always use try/finally for session cleanup
2. Set timeouts as safety nets against runaway sessions
3. Use cost thresholds on all production AI operations
4. Monitor dashboard regularly for unexpected usage patterns
5. Test with small batches before scaling up operations