Prompt Token Optimizer
Trim filler and repeats from long prompts, review every change, and see what it saves each month.
Suggested Trims
Safe: spacing and wordy phrases. Likely safe: filler that rarely changes meaning. Review: can change emphasis or intent, so it is off by default.
Review the Changes
Red is removed, green is added. Click any change to switch just that one on or off. Dashed outlines are available but off. A dot (·) is a space and ↵ is a line break.
Shorter Prompt
Monthly Savings
Send to Cost ForecasterThe forecaster opens with your trimmed size and shows the monthly saving against the original, using your own volume and caching settings.
What It Is For
Long prompts are re-sent on every request, so every wasted word is paid for again and again. This tool finds filler, repeats and wordy phrases, shows each change, and lets you decide what to keep. Then it turns the token saving into a monthly figure in the Cost Forecaster.
What to Expect
Rule-based trimming is modest. On verbose, human-written prompts, savings of roughly 10 to 30 percent are typical. Bigger cuts usually need a human to merge overlapping instructions or shorten examples.
Questions
Will the trimmed prompt behave the same?
Not guaranteed. The tool never changes numbers, links, {{variables}}, quoted text, code, tags, or anything inside tags such as <example> and <document>, and it checks that they are all still there. But wording can matter to a model, so test the shorter prompt on a few real inputs before switching.
Why are some trims off by default?
Words like "important", "make sure" and "very" can signal emphasis. Removing them may weaken an instruction, so they are off until you choose.
Is my prompt uploaded?
No. Everything runs in your browser. Nothing is sent anywhere.
Why might the saving be smaller than expected?
If your system prompt is cached, the repeated part is already billed at a reduced rate, so shrinking it saves less. The Cost Forecaster models this when you turn caching on.