These fall into three groups. If you are here because a prompt is not working, start with the Prompt Optimizer or the Quality Checker. The first rewrites a prompt into the five-part structure that closes the gaps models fill with guesswork; the second grades one and tells you specifically what is missing rather than giving a verdict.

If you are here because something costs too much or will not fit, use the Token Counter, Cost Calculator and Context Window Calculator. Tokens are the unit models charge for and run out of, and both problems are the same measurement viewed from different sides.

If you are building something, the System Prompt Builder, JSON Formatter and Prompt Splitter cover the three things that break most often: inconsistent behaviour across a conversation, structured output that will not parse, and documents longer than the interface accepts.

Why they run in your browser

Every tool here is client-side JavaScript. Nothing you paste is uploaded, logged or stored — there is no server that could receive it. That is a deliberate constraint rather than a feature we are advertising: people paste real work into a token counter, and a tool that quietly transmitted a draft contract or a block of production code to count its characters would be a bad trade for everybody.

It also means there is nothing to sign up for, no rate limit, and no account to lose. Close the tab and nothing about what you did survives.

Where they are approximate

The token-based tools estimate rather than tokenise exactly, landing within roughly 10–15% for English prose and less reliably for code and other languages. That is fine for planning — deciding what fits, comparing models, sizing a budget — and not fine for billing reconciliation or for sitting on the edge of a hard context limit. Each tool page says where its own limits are, and when to use your provider's official tooling instead.