Tokenline

tokenline
See your AI coding costs in real time. Tokenline adds context usage, prompt-cache savings, TTL countdown and rate-limit pacing to Claude Code and Gemini CLI.

Quickstart

1. Install (via npm)

Run the installer for your AI coding CLI:

Claude Code Per-project installation (./.claude/settings.json)

Claude Code Global installation (~/.claude/settings.json)

Antigravity CLI Global installation (~/.gemini/antigravity-cli/settings.json)

2. Restart your CLI

Simply restart Claude Code or Antigravity, and your new statusline will be live.

Updating

The installed tokenline.sh is a copy, so it stays on the version you installed until you update it. Use the same scope flag you installed with:

update finds the script your statusLine points at and replaces it. It never edits settings.json, and there's nothing to restart.

Because of that, a feature that needs a new settings key isn't switched on by update alone. Installed before subagent rows? update says so and prints the one command that adds them, for the same scope and --dir, e.g. npx -y @inbrace-tech/tokenline@latest init --global. Re-running init is safe: it confirms your existing statusLine and only adds subagentStatusLine.

You don't have to remember to check. An installed copy looks up the latest published version in the background: when you open a new session (if the last check is over an hour old), and at least every 6 hours in a long one. When a newer version exists, it adds one gray line to the statusline:

In Claude Code, type that line (with the leading !) at the prompt to run it without leaving the session. The next refresh shows the new version.

The check is a single GET to registry.npmjs.org for the version number, made only if curl is installed. It runs detached, so rendering never waits on the…