Skip to main content
Costs vary widely by model and provider, and a single expensive workflow can quietly drive up your bill without anyone noticing. Monitoring aggregates usage as it happens so you can catch spikes early instead of finding out at invoice time.

​
What Monitoring Shows

Monitoring is a project-level dashboard that tracks:
  • Token Consumption - Input and output tokens by LLM model
  • Cost Breakdown - Costs by provider (Claude, GPT-4, Perplexity, etc.)
  • Service Usage - Resource consumption by tool type In Roadmap
  • Trends - Usage patterns over time In Roadmap
  • Budget Tracking - Spending vs budget limits In Roadmap
Monitor shows CONSUMPTION and COSTS. For detailed analytics about agent performance, conversations, and execution logs, see Agent Analytics.

​
Cost Tracking

​
Understanding Your Costs

Cost structure:
  • LLM Services (Claude, GPT-4, etc.)
    • Based on tokens consumed
    • Different rates for input vs output
    • Rates vary by model
  • Tool Services (Web Search, Code Execution, etc.)
    • Based on usage (searches, executions, calls)
    • Per-unit pricing or flat fees
    • Varies by service
  • External Services (via API tools)
    • Billing by external provider
    • PLai may charge platform fee
    • Check service docs for pricing

​
Cost Per Agent/Workflow

Track which agents or workflows are most expensive: Estimated Cost Breakdown:
These are estimated breakdowns. For exact per-agent costs and usage details, check Agent Analytics.

​
Reducing Consumption

​
Optimize Token Usage

Savings: 77% reduction

​
Best Practices

Concise Prompts

Every token costs. Be brief and specific.

Smart Context

Include only relevant information, not full datasets.

Cheaper Models

Use Haiku for simple tasks, Sonnet for complex.

Batch Processing

Process multiple requests together when possible.

​
Example Optimization

Email Support Bot:
  • Problem:
    • Current: 500 requests/day
    • Avg input tokens: 1,500 per request
    • Daily token cost: 2.25(500×1,500×2.25 (500 × 1,500 × 3/1M)
  • Optimization:
    • Reduce context from 1,500 to 500 tokens
    • New daily token cost: 0.75(500×500×0.75 (500 × 500 × 3/1M)
    • Monthly savings: $45 (66% reduction)

​
Troubleshooting

Steps:
  1. Check which model/service increased
  2. Look at token consumption
  3. Review recent agent/workflow changes
  4. Optimize expensive components
  5. Set alerts to catch next time
Check:
  • Monitor refreshes every 5-15 minutes
  • Some costs finalized at end of day/week
  • Different providers have different billing cycles
  • Wait a few hours for full update
Verify:
  • Budget is set in project settings
  • Alert threshold configured correctly
  • Notifications enabled
  • Check spam folder for alert emails
Use:
  • Agent Analytics for per-agent data
  • API endpoints for programmatic access
  • Export data for custom analysis

​
Monitoring Best Practices

Review Weekly

Spend 5 minutes weekly checking consumption

Set Budgets

Know your spending limits before issues arise

Enable Alerts

Get notified of cost spikes immediately

Optimize Early

Don’t wait for problems. Optimize as you build.

​
Monitoring Checklist

  • ✅ Check project monitoring dashboard daily
  • ✅ Review cost trends weekly
  • ✅ Set realistic budget limits
  • ✅ Configure alert thresholds
  • ✅ Identify expensive agents/workflows
  • ✅ Optimize high-consumption components
  • ✅ Export reports for stakeholders
  • ✅ Archive data per retention policy

​
Key Takeaways

✅ Monitor tracks consumption - Tokens and service usage
✅ Understand costs - Different models and services cost differently
✅ Set budgets & alerts - Avoid surprise bills
✅ Optimize early - Reduce consumption before it becomes expensive
✅ Track trends - Understand patterns and plan capacity