Skip to content

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

5 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

burnr 🔥

A bonfire of glyphs in your terminal, burning the tokens your Claude Code session burns.

burnr attaches to your current project's live Claude Code session transcript and turns every turn's token usage into a burst of rising, fading Matrix-style embers. Big turns blaze; quiet sessions smoulder down to a gentle flicker. A status line keeps count.

burnr in action

Install

cargo install burnr

Usage

Run it from the project directory whose Claude Code session you want to watch:

burnr

It finds the latest session transcript under ~/.claude/projects/, tails it live, and rotates automatically when you start a fresh session. Press q to quit.

Run it in a tmux/terminal split beside Claude Code and watch your conversation burn.

Options

burnr                        # watch the current directory's latest session
burnr --project <dir>        # watch another project's latest session
burnr --session <path>       # pin to one specific transcript (.jsonl)
burnr --demo                 # no ingestion; press space to fire synthetic bursts

Status line

burnr — burned 1.2M · last 88.3k — q to quit

burned is the cumulative token count for the watched session — including cache reads, which the model re-pays on every turn, so it grows much faster than your context size. last is the most recent turn.

Configuration

Optional, at ~/.config/burnr.toml:

[idle]
after_seconds = 10.0             # settle to a flicker this long after the last burst
embers_per_second = 4.0          # gentle idle flicker rate
active_embers_per_second = 30.0  # ambient smoulder rate while the session is active

Every field is optional; unknown fields are ignored.

How it works

Claude Code writes a JSONL transcript per session under ~/.claude/projects/<slugified-project-path>/. Each assistant message carries a usage snapshot (input, output, cache-read and cache-creation tokens). burnr tails the newest transcript, de-duplicates streaming snapshots, and maps each turn's total through a log-scaled intensity curve — calibrated against a few thousand real turns — into particle count and rise speed.

Licence

MIT

About

Token burning tui visualisation

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages