Rashed
PromptMeter is a free browser tool that counts tokens and turns a loose prompt into a structured brief, before you pay an API to discover the mess.
Local processing
The prompt stays in this tab. Counting, scoring, and the rewrite do not upload it, and the page does not save it in the browser.
Estimate, not a bill
Token counts aim at cl100k_base. The three prices are public list rates for input only, dated on the tool. They are not an invoice.
Open methodology
The density score starts at 40 and moves on a fixed list of signals. The guide on the home page lists every rule.
The site exists because a prompt box hides both problems. A polite paragraph can contain no task. A short paragraph can still be hundreds of tokens once you paste a policy, a schema, and three examples. The counter estimates cl100k_base, the tokenizer family used by GPT-4-class models, inside the browser. The density score is a checklist: filler phrases, action verbs, labels, delimiters, constraints, and examples. The optimizer rearranges a draft into Role, Task, Context, Constraints, Output format, and Examples without calling a model.
Rashed builds and edits the pages. The notes are written for people who pay for API calls or who paste the same brief into a chat every day and want it to stay stable. They are not a course, and they are not affiliated with OpenAI, Anthropic, or any other model provider. Prices mentioned on the tool are public list prices dated on the page. They are illustrations.
The tool does not have accounts. A prompt you paste is not uploaded and is not stored in the browser. Corrections are welcome. A token estimate that drifts, a price that has gone stale, or a note that overclaims should be fixed in the open. The point of the site is a number you can check and a brief you can read, not a score that flatters the draft.