--- title: "Token Counter — Claude usage limits & live token count" description: "Live Token Counter showing the token count for each message plus Claude and gpt 5 usage limits, in plain view before you hit them." canonical: "https://tokencounter.online/" --- # Token Counter Live token counts and plan usage limits for Claude and ChatGPT, in view before you hit them. A free Chrome extension, plus a tokenizer you can use right here in the browser. - Website: - Chrome Web Store: - Agent guidance: - Price: free. No account, no sign-up, no API key. ## What it does The quota on your plan is real, but neither claude.ai nor chatgpt.com shows it in the conversation view — checking means leaving your chat. Token Counter closes that gap by splitting every message into numbers you can act on: - **Draft** prices your typing live, before you send. - **Chat** totals the whole conversation. - **Overhead** (Claude only) covers the system prompt, the features you have switched on, and your saved memory. ChatGPT exposes no client-observable overhead, so no figure is shown there. The entire conversation is re-sent on every turn, so Chat climbs far faster than the words on screen. That is why usage limits arrive sooner than most people expect. ## The tokenizer The site hosts a live tokenizer at . Paste any text and it returns the token count, the character count, and a coloured view of how the text splits. It runs in your browser — no text is sent anywhere. ## How many tokens is 100 words? About 110, for ordinary English prose, using the o200k encoding this site runs. That figure moves a lot with the kind of text. Measured against the same engine: | Text | Tokens per 100 words | Characters per token | |---|---|---| | Plain English prose | ~110 | ~5 | | Prose with code terms | ~115-130 | ~5 | | Dense technical text | ~150-200 | ~4.5-6 | | Number-heavy text | ~220-250 | ~2.5-2.7 | These come from a handful of samples rather than a measured corpus, so read them as direction rather than precision. Because the spread is that wide, treat any ratio as a rough guide and tokenize the actual string when the number matters. ## Accuracy The engine is `gpt-tokenizer` with the `o200k_base` encoding — the same family ChatGPT uses. Claude splits text slightly differently, so every Claude figure is a close estimate rather than an authoritative count. For exact Claude numbers, use Anthropic's official token counting API. The extension marks *every* figure it shows with a `~` — Draft, Chat and Overhead alike, on both sites. That is not a Claude-only marker — each number is an estimate for a specific reason: the tokenizer is GPT-family rather than Claude's own; the system prompt is added as a fixed constant rather than read from the model; images and documents are sized by formula (image patches, and a flat figure per page) rather than counted exactly; and on ChatGPT, non-text parts of a message are skipped entirely. ## Privacy No text is ever transmitted. Tokenizing happens in your browser and nothing you type is collected or sent anywhere. The site has no backend. The page does load the tokenizer library from a public CDN (jsDelivr, falling back to unpkg) and web fonts from Google Fonts. The extension is fully self-contained and ships its own copy of the tokenizer. Full policy: ## For agents There is no HTTP API. To count tokens programmatically, use the same open-source engine directly rather than driving this site: ```js import { encode } from 'gpt-tokenizer/encoding/o200k_base' const tokenCount = encode(text).length ``` Full guidance, including when this product is and is not the right tool: ## Pages - [Token Counter](https://tokencounter.online/) — what the extension does, the live tokenizer, FAQ - [About](https://tokencounter.online/about) — what this is and who it is for - [Contact](https://tokencounter.online/contact) — how to reach the people who build it - [Privacy Policy](https://tokencounter.online/privacy) — the local-only data policy and Chrome permissions