Your OpenAI bill landed. It's 3x what you expected. You open the usage dashboard and trace the spike back to three days ago — a retry storm that ran all night while everyone was asleep.
This is how it happens for most teams. Not through negligence. Through the fundamental problem with passive monitoring: by the time you see the chart, the damage is done.
Spend alerts flip this. Instead of checking dashboards, you get interrupted only when something is wrong. Here's how to set them up in under 5 minutes.
Get weekly AI cost optimization tips
Join engineers cutting their AI bills. No spam — just strategies that actually move the number.
The 4-Step Setup
Sign up at SpendPilot
Go to spendpilot-2.polsia.app/login.html and create a free account. No credit card required. You'll be inside the dashboard in under a minute.
Connect your OpenAI or Anthropic API key
Navigate to the API Keys section and add your key. SpendPilot uses it read-only to pull billing data from the provider's usage API — it can't make API calls on your behalf. Keys are stored encrypted with AES-256-GCM. Once connected, SpendPilot immediately imports your last 30 days of spend history, broken down by model and day.
Set a spend threshold
Go to Alerts and create a new rule. Pick a provider (OpenAI, Anthropic, or all providers combined), enter a dollar amount, and optionally set an email to notify. That's it. A good starting point: set your threshold at 80% of your monthly budget. If you spend $500/month on OpenAI, set an alert at $400.
Get alerts before you overpay
SpendPilot checks your thresholds every time new spend data syncs. The moment cumulative spend crosses your threshold, you get an email with the provider, the exact amount, and the timestamp. No dashboards to check. No Slack channels to monitor. You get interrupted when it matters and left alone when it doesn't.
If you're not sure where to start: look at your last 3 months of API spend, take the highest month, and set your alert at 90% of that number. You'll catch runaway usage before it exceeds your worst-case month. Adjust from there once you have a sense of your normal baseline.
Why Email Alerts Beat Dashboard Monitoring
Dashboards require you to remember to check them. Engineers don't do that — not reliably, not at 2 AM when a retry storm starts, not during a sprint when everyone's heads-down on a deadline.
Email alerts are push, not pull. The system watches your spend continuously. You get notified exactly when something breaches your threshold, with enough context to act: which provider, which amount, when it happened.
"We stopped checking the OpenAI dashboard entirely. If SpendPilot hasn't emailed us, spend is under control. That's the whole workflow."
The alert history log also gives you an audit trail — useful when finance asks why the API bill was higher in March, or when you're doing a post-mortem on an incident.
Common Threshold Mistakes
Setting the threshold too low. You'll get an alert every week and start ignoring them. Alert fatigue is real. Set thresholds at meaningful levels that signal actual problems, not normal variance.
Setting only one threshold. Consider setting per-provider thresholds in addition to a combined threshold. An OpenAI-specific alert catches model routing issues (e.g., agents calling GPT-4 when they should use GPT-3.5). A combined threshold catches total spend creep.
Not updating thresholds as you scale. When you double your agent fleet, your normal spend doubles too. Review your thresholds when you make significant changes to your AI infrastructure.
What Happens After You're Set Up
Most teams configure alerts once and never touch them again. The dashboard is there when you want a detailed breakdown — spend by model, daily trends, alert history — but the day-to-day workflow is just: build things, get interrupted if spend spikes, investigate, fix it.
That's the point. Your job is building products, not monitoring API costs.
Want context on why spend controls matter at scale? Read Why Your AI Agents Need a CFO or 5 Signs Your AI Agent Fleet Is Bleeding Money.
See how SpendPilot compares to Portkey, Helicone, and other AI cost tools.
Not sure what threshold to set? Use the AI API cost calculator to estimate your expected monthly spend and set a realistic baseline.
Once alerts are in place, 7 cost optimization strategies covers how to systematically reduce your baseline spend.
Curious what AI agents actually cost to run? The Real Cost of Running AI Agents in 2026 breaks it down by model, provider, and scale.