Usage Bar for Claude

See your Claude plan usage, context window and prompt-cache countdown on claude.ai — as a top bar or inside the chat box.

See your Claude usage limits, token usage and prompt-cache timer, directly on claude.ai. Usage Bar for Claude adds a compact, native-feeling usage indicator to the Claude.ai website so you always know how much of your plan limits and context window you've used, without leaving the conversation. New in 1.0.0: a rebuilt context estimator and a token breakdown. The counter now reads the full conversation data, follows the active branch of the chat, adapts the window size to your plan and model (200k, 500k or 1M), and warns you when it is time to start a new chat. Hover the ring for a hollow pie chart showing exactly where your tokens go: your messages, Claude's replies, thinking, tools, and files. Features: - Usage bar showing your 5-hour limit as a percentage with a live "resets in…" countdown, pinned to the top of the page or as a slim line inside the chat box (pick the placement in the popup) - Hover panel with full plan usage: 5-hour limit, weekly all-models limit, extra credits, and Routines - Context-window token counter: a circular indicator estimating how much of the 200k context window the current chat occupies, distinguishing the active conversation branch from total tokens used (edits and retries don't inflate the count) - Smarter token estimation in 1.0: reads the full conversation data, including tool calls and attachments the page never shows, with a script-aware heuristic (word-based for spaced languages, character-based for Chinese/Japanese/Korean) - Prompt-cache countdown after each assistant reply, so you know when a follow-up still hits the ~90% cheaper cache - Available in 14 languages: English, Español, Português (Brasil & Portugal), Français, Deutsch, Italiano, Русский, 简体中文, 繁體中文, 日本語, 한국어, हिन्दी, Türkçe - Performance-friendly: all polling and rendering suspends while the tab is hidden Privacy first: - Uses your existing authenticated claude.ai session; only same-origin requests, never a third-party server - The only permission is "storage", no host permissions, no analytics, no telemetry - Stores only sanitized usage metadata and numeric estimates; conversation text is processed in memory only. No prompt, response, cookie, or auth data is ever stored - Open source (MIT): github.com/disi910/claude-usage-bar About the context indicator: Claude's API does not expose token counts, so the extension estimates them locally from Claude's own conversation data using a calibrated, script-aware heuristic. The system prompt and tool definitions aren't visible to any extension. Treat the percentage as a close indicator, not a guarantee. If you run into a bug or have a feature request, open an issue on GitHub or email me and I'll take a look.