All tools

Claude Tokenizer

This tokenizer counts tokens in your CLAUDE.md, system prompt, or any text before you send it to Claude. See how your content fits in the context window — runs entirely in your browser, no data sent anywhere.

Your text
0 characters0 words

Estimated tokens

0

≈ 4 chars per token (English prose)

Context window usage

Claude Haiku 4.5Fastest & cheapest
0.0%
0 / 200,000 tokens
Claude Sonnet 4.6Best balance
0.0%
0 / 200,000 tokens
Claude Opus 4.6Most capable
0.0%
0 / 200,000 tokens

Tips for reducing tokens

  • Remove redundant comments and examples
  • Use bullet points instead of prose
  • Keep code snippets short and focused
  • Browse lean community rules for inspiration

Find lean CLAUDE.md rules

Browse community-optimized rules that balance context and effectiveness.

Browse rules

What is a Claude token?

Claude uses a byte-pair encoding (BPE) tokenizer similar to GPT models. Roughly speaking, one token is about 4 characters of English prose or 3 characters of code. A word like "tokenization" counts as 2–3 tokens. The Claude API bills by token — both input (prompt) and output (response).

What is a context window?

The context window is the maximum number of tokens Claude can process in a single request. Claude Haiku 4.5, Sonnet 4.6, and Opus 4.6 all support 200,000 tokens (~150,000 words). If your prompt plus expected output exceeds this limit, you will need to shorten your input.

Tips for reducing token count

  • Remove redundant comments and examples from CLAUDE.md
  • Use bullet points instead of prose paragraphs
  • Keep code snippets short and focused — show patterns, not full implementations
  • Avoid repeating the same instruction in multiple ways
  • Browse lean community rules for inspiration