The $800 Weekend Problem
It happens more often than anyone admits: a developer deploys an agent on Friday, a prompt is misconfigured or a retry loop kicks in, and by Monday the OpenAI bill shows $800 in unexpected charges. No one knew. No one was alerted. The damage was already done.
Budget alerts are the single most valuable feature in ClawHQ. They're the difference between "we caught it in 5 minutes" and "we caught it on the invoice."
Types of Budget Alerts
1. Per-Agent Daily Budget
Set a maximum daily spend for each agent. When the agent approaches or exceeds this limit, you're notified immediately.
Example: Your research agent normally costs $5/day. Set a budget of $10/day. If a bad prompt causes it to spike to $8, you get a warning. If it hits $10, you get a critical alert (or the agent pauses automatically).
2. Cost Spike Detection
Alerts when any agent's hourly spend exceeds a multiple of its recent average. This catches anomalies without requiring you to set fixed thresholds for every agent.
Example: Agent normally spends $0.50/hour. If it suddenly hits $3/hour (6x spike), you're notified immediately.
3. Monthly Projection Alert
Based on current spending trends, ClawHQ projects your end-of-month total. If the projection exceeds your budget, you get warned with weeks to spare.
Example: You budgeted $500/month. On the 10th, ClawHQ projects $750 at current rates. You get an alert and have 20 days to optimize.
4. Team Budget
Aggregate budgets for groups of agents. Essential for organizations where different teams have different AI budgets.
5. Global Fleet Budget
A backstop for your entire fleet. If total spending across all agents approaches your global limit, everyone with alert access is notified.
Setting Up Alerts in ClawHQ
Navigate to Dashboard → Alerts → Create Rule:
- Choose scope: Per-agent, per-team, or global
- Set threshold: Dollar amount or percentage of budget
- Choose severity: Warning (80% of budget) or Critical (100%)
- Configure action: Alert only, throttle, or suspend
- Set notification channels: Email, Slack, Discord, webhook
Recommended Alert Configuration
Here's the alert setup we recommend for every team:
- Per-agent daily budget: 2x the agent's average daily spend
- Cost spike detection: Alert at 3x hourly average
- Monthly projection: Warning at 80% of monthly budget, critical at 100%
- Global fleet budget: Your total monthly AI budget with 80%/100% thresholds
Alert Actions: Beyond Notifications
On ClawHQ Pro and above, alerts can take automated actions:
- Alert only: Send notification, agent continues running
- Throttle: Reduce the agent's request rate to slow spending
- Suspend: Pause the agent entirely until a human reviews
For production agents, we recommend "alert + throttle" as default and "suspend" for non-critical agents.
Real Stories: How Alerts Saved Teams Money
- A startup caught a $400/day runaway agent within 15 minutes — saving an estimated $2,800 that week
- An agency noticed one client's agents were costing 4x more than estimated, allowing them to adjust pricing before losing margin
- A solo developer's budget alert caught a retry loop that would have consumed their entire monthly API budget in 3 hours
What Alerts Won't Do
Alerts are reactive — they catch problems after they start. For proactive cost management, combine alerts with:
- Model optimization — use cheaper models where possible
- Prompt optimization — reduce token waste
- Response caching — avoid redundant API calls
