Billing & Usage for AI Agents

The Billing & Usage page is your control center for understanding your org's Agent and Data App usage in NinjaCat — what's being used, who's using it, and how much credit you have left.

🔒
Who can see this page?
Billing & Usage is visible only to users with the Manage Agency Settings permission toggle enabled for their assigned User Role.

The page has two tabs: AI Usage and AI Billing.

AI Usage Tab

The AI Usage tab is the landing page under Billing & Usage. It shows you the actual activity happening across your agents — measured in raw tokens and actions.

Here's what you'll find:

Category	What it measures	Segment / breakdown options
AI Tokens	Tokens consumed by LLM models processing your agents' requests (prompt + completion tokens). Every model has its own rate per token.	• By Provider (e.g. Anthropic, OpenAI, Google) • By Model (e.g. Claude Sonnet vs. Claude Opus) • By Token Type (Prompt vs. Completion) • By Product (Agents vs. Data Apps)
AI Actions	Non-chat agent activity — generally tool use, such as web scrapes, image generation, web research, and more.	• By Action (image_generator, web_scraper, google_web_search, researcher, query_executor, browser_use, image_editor, video_generator) • By Product (Agents vs. Data Apps)
Data Compute	Data warehouse compute consumed when your agents query and process data.	Additional breakdowns coming soon

Understanding the Top Usage Table

There are two ways to slice the token and action data in this table:

Top Usage by Agent — token consumption and actions broken down by Agent Name
- Any built Agent that has consumed tokens will populate here based on the date range selected
- "Bob" refers to the Agent Builder Bob and its token consumption
- "Appy" refers to the Data App Builder and its token consumption
- "Other" was capturing token consumption for tokens that were not being attributed to an Agent prior to June 2nd. Changes were made on that date so that all tokens are now being attributed to an Agent, but those changes could not be made retroactively. When viewing data after June 2nd, you will not see an "Other" line.
Top Usage by User — token consumption and actions broken down by User
- "Other" captures tokens that weren't executed directly by a human user. This includes scheduled and webhook triggers — set up by a human user initially, but then initiated automatically later according to the trigger's rules — and when an agent is calling on a subagent

AI Billing Tab

The three credit categories in the AI Billing tab map directly to the three categories of raw usage tracked in the AI Usage tab:

AI Usage (raw)		AI Billing (credits)
AI Tokens	→	AI Token Credits
AI Actions	→	AI Action Credits
Data Compute	→	Data Compute Credits

The AI Billing tab translates that activity into credits so you can manage your plan. Your plan that you have with NinjaCat is based on these credits. In order to use our Agents and Data Apps features, there needs to be available credits.

This tab shows:

Remaining credit balance out of your total number of credits purchased
Estimated days remaining
- This estimate looks at your credit usage over the past 90 days, giving more weight to recent activity and accounting for whether usage is trending up or down. We project that pattern forward to estimate when your remaining credits would run out.
- The headline number reflects your typical usage rate. The listed range of days shows how the estimate shifts if usage runs higher than usual (the low end) or lower (the high end). When usage varies a lot, your typical rate can last beyond a year (365+) while a busier-than-usual stretch could still use credits sooner.
Estimated depletion date — the day your credits are projected to run out based on the above information
Credit category breakdown — how your credits are being spent across Token, Action, and Data Compute credits
Credit transaction history — credits purchased, credits spent per day, and end-of-day balance

Use this tab when you want to answer "How much have we spent and how long will our credits last?"

Avoiding a lapse in access when credits run low

When credits run out, agent requests will start to fail. To avoid that:

Reach out to your NinjaCat account manager or customer success manager to purchase additional credits.
Set a daily credit limit (Admins only). This caps how much of your overall plan can be burned in a single day, which protects you from accidentally draining your full balance in one busy day.
Watch the estimated days remaining and depletion date on the AI Billing tab — these update based on your actual burn rate and give you early warning.

Credit Definitions

NinjaCat uses a single credit currency:

100 credits = $1.00 (1 credit = $0.01)

There is no fixed credit-to-token ratio — the exact cost depends on the model, provider, and action being used. Credits are grouped into three categories:

Category	What it covers
AI Token Credits	Credits consumed by LLM models processing your agents' requests (prompt + completion tokens). Every model has its own rate per token. A single agent run will typically draw from both the AI Token Credits and AI Action Credits buckets, as it typically includes chat-based 'thinking' time + some variation of tool calls.
AI Action Credits	Anything that's non-chat agent activity — generally tool use. This covers the built-in tools an agent uses, such as web scraping, image generation, web research, and data queries. The following built-in tools count as AI Actions and generate AI Action Credits: • Image Generator (`image_generator`) • Image Editor (`image_editor`) • Video Generator (`video_generator`) • Web Scraper (`web_scraper`) • Google Web Search (`google_web_search`) • Researcher (`researcher`) • Query Executor (`query_executor`) • Browser Use (`browser_use`) Custom Actions and MCP Server Tools are not included in this category. Different actions are priced differently — some charge per invocation, some by browser time, some by units consumed (like Snowflake credits). It all gets converted to dollars and then to NinjaCat credits at the standard 100-credits-per-dollar rate.
Data Compute Credits	Credits consumed by data warehouse compute when your agents query and process data.

A few things worth knowing

Builder Bob (NinjaCat's in-platform agent builder assistant) shows up in Top Usage by Agent and does count against your token usage.
Data Apps are included in credit consumption and shown as a separate line item.
The AI Insights Widget in reporting is not part of AI credit pricing — it's covered under your reporting package.