Settings

Usage data

Monitor Flow executions, AI token usage, and costs to understand and optimize your Runtype spending.

Accessing usage data

  1. Go to Settings → Usage
  2. View usage analytics with charts and breakdowns
  3. The Usage tab defaults to the current Billing month. Use Period to switch between Billing month, 7-day, 30-day, or 90-day views
  4. Use Refresh to request the latest aggregate or Export CSV to download the same analytics shown on the page

Key metrics

The Usage tab opens with a row of summary tiles for the selected period: Runtype-billed usage, executions, average cost per execution, and estimated cache savings with cache hit rate (the cache tile appears once your usage has prompt-cache activity). Below the tiles are chart and table views:

  • Usage cost over time: track cost trends
  • Executions over time: monitor execution volume
  • Usage cost by category: how cost splits between AI models and built-in tools
  • Cost by AI model: which models drive the most cost, including what share of each model’s input tokens was served from the prompt cache
  • Cost by built-in tool: tool-level cost attribution
  • Prompt cache: cache read and write token totals. This card appears once your usage has cache activity. See Prompt caching for how caching works.

An execution is one Flow run, batch-record run, Eval run, or scheduled trigger. The totals, charts, breakdowns, and CSV export use the same usage report. Recent activity can take a short time to appear. The page shows when it was last updated and warns when the selected period begins before the available history; missing coverage is not proof of zero usage.

Usage refreshes automatically and can also be refreshed manually. Settings → Billing shows the current amount toward your monthly cap. It can be slightly ahead while recent activity is still appearing on Usage, and it determines when new executions pause. Your issued invoice shows the amount charged to your card.

Cost optimization tips

Use cheaper models when possible

GPT-5.4 Nano is significantly cheaper than GPT-5.4 for simple tasks.

Reduce max tokens

Lower max tokens limits prevent unnecessarily long responses.

Cache API responses

Store results in Records instead of making duplicate API calls.

Batch processing

Process multiple items per execution to amortize overhead costs.

Monitor and optimize prompts

Shorter, more focused prompts reduce input token costs.

Understanding charges with your own keys

When using BYOK:

  • Runtype estimates costs based on token usage
  • Actual charges come from providers (OpenAI, Anthropic, etc.)
  • Check provider dashboards for precise billing
  • Runtype’s estimates help predict costs but aren’t invoices

Next steps