Count tokens, track context fill, and get trim suggestions the moment you paste a prompt or drop a file.
Pick a model, paste a prompt, or drop a file and the page updates the token view live. That makes it easy to see how close you are to the context window before you send anything.
Some model families can be counted more precisely than others, so the tool labels the result as exact or approximate instead of pretending every number has the same confidence.
When a prompt goes over budget, the Trim block points at obvious places to cut first. Everything stays in your browser, and file mode is for quick sizing rather than deep document extraction.
