Paste your prompt, pick a context window size, and see how much of it you're using.
No — this counts only the prompt you paste. Most APIs count input and output tokens against the same context window, so leave headroom for the response.
No. Everything happens locally in your browser after a one-time tokenizer download.