Token Calculator.
What does a single AI request cost you, and what does a whole agentic application cost? The Token Calculator gives answers to both questions, locally in the browser. In the Agentic Forecaster you build complete workflows by setting functions, frequencies and models, and read off day, month and year cost. Tab two shows what a token actually is and why the same content costs differently across languages. Tab three takes a single prompt and runs it in parallel across all relevant providers. Everything runs in the browser, with no API key. Your input never leaves the window.
Open the tool→Four moments to reach for this tool.
Before API integration
Before going live with an AI feature, the typical token load is worth a look. A handful of realistic inputs turn into a defensible cost curve.
Stakeholder briefing
The question "what does it cost per request?" is best answered live in tab 1. The question "what does it cost in total per month?" is what tab 2 does.
Choosing a model
A frontier model is not always the right pick. The model-comparison heatmap in the forecaster shows where a workhorse or budget tier is enough.
Cloud vs on-premise
When volume or confidentiality demand it, the forecaster compares cloud monthly cost against GPU server capex plus power and maintenance. With break-even hint.
Bring your scenario into the tool.
Three tabs for three views on the same numbers: cost per function, projection across frequency, and the comparison between frontier and budget models. The hints right under the tabs adapt to whichever tab is active.
Insights that shape architecture and budget downstream.
Build your own setup.
Want to save your own scenarios, compare future models when pricing shifts, or extend your agent setup and recalculate without starting over? Set up a free QCT account. Your setup stays put, you can adjust and extend any time, and you can have it mailed as a structured summary to yourself or your team.
Want to go deeper into tokens and APIs? That happens in the intensive.
Two days hands-on with Claude and ChatGPT, working through token streams, context windows, API limits and cache strategies in detail. You leave with a clear sense of which dials to turn in a real project when costs start to drift.
To the intensive course→