CostPilot AI

Inspiration

AI teams track performance well, but lack visibility into API costs.
As usage scales across projects, costs become hard to attribute — leading to unexpected losses.
CostPilot AI was built to make AI spending transparent and controllable.

What It Does

CostPilot AI helps teams:

• Track token usage per project
  
• Calculate profit margins in real time
  
• Monitor burn rate (🟢 <60%, 🟡 60–85%, 🔴 >85%)
  
• Compare models to find cheaper alternatives
  
• Automate logging via API/webhooks
  

How We Built It

• Relational model: Projects + Models + Usage Logs
  
• Dynamic cost calculation from token usage
  
• Burn rate indicator system (60% / 85%)
  
• Model comparison engine for cost simulation
  
• Webhook API for automated logging
  
• Token estimation (multi-method approximation)
  

Challenges We Ran Into

• Making cost data actionable, not just visible
  
• Handling incomplete or inconsistent usage data
  
• Keeping calculations consistent across all views
  

Accomplishments

• Built a working model comparison engine for real savings insights
  
• Designed an intuitive burn rate system
  
• Implemented flexible token estimation methods
  
• Delivered full API integration workflow
  

What We Learned

• Cost visibility directly influences engineering decisions
  
• Simple UX (like color indicators) can outperform complex analytics
  
• Financial tracking should be built into systems, not added later
  

What’s Next

• Predictive budget forecasting
  
• Team collaboration (roles & permissions)
  
• Custom alert thresholds
  

Built With

• medo
• react
• supabase
• typescript
• vite