Tools
Small utilities for working with AI.
Free, no signup, no tracking. Built because we kept needing them internally and figured we'd share.
All three answer questions that come up in the first hour of scoping an AI project: how much text is this really, what will it cost to run every month, and is the thing it replaces expensive enough to bother. They're estimators, not billing systems — close enough to decide with, not close enough to invoice on.
Token counter
Estimate how many tokens a prompt or document will use. Approximation good enough for scoping; not byte-perfect.
OpenLLM cost estimator
Monthly cost for a conversational AI workload — pick model, set traffic shape, see the spend. Built from Anthropic + OpenAI's public pricing.
OpenROI calculator
Estimate the dollar value of a deflection / automation use case. Already on /roi — linked here for convenience.
Open
What these numbers are good for
- Token counts are approximations, and that's usually fine
- Every model family tokenizes differently, so an exact count only exists for one specific tokenizer. For scoping, the number you need is the order of magnitude — whether a document is 2k tokens or 200k changes the architecture; whether it's 2,000 or 2,150 changes nothing.
- Cost is driven by traffic shape, not model choice
- Teams tend to agonise over per-token pricing and then miss that their real cost driver is conversation length and how much context gets re-sent on every turn. Change the traffic assumptions in the estimator before you change the model — the swing is usually larger.
- ROI is the number that kills projects, early or late
- A deflection use case that saves four minutes on a ticket nobody logs is not a project. Running the ROI calculator before the build is how you find that out cheaply instead of at the retro.
Want a second opinion on the numbers you get? Book a 20-min scoping call →
