cburn/README.md

cburn

A CLI and TUI dashboard for analyzing Claude Code usage metrics. Parses JSONL session logs from ~/.claude/projects/, computes token usage, costs, cache efficiency, and activity patterns.

Installation

go install github.com/theirongolddev/cburn@latest

Or build from source:

git clone https://github.com/theirongolddev/cburn.git
cd cburn
make install

Requires Go 1.24+.

Note: Ensure ~/go/bin is in your PATH:

# bash/zsh: add to ~/.bashrc or ~/.zshrc
export PATH="$HOME/go/bin:$PATH"

# fish: add to ~/.config/fish/config.fish
fish_add_path ~/go/bin

Quick Start

cburn              # Summary of usage metrics
cburn tui          # Interactive dashboard
cburn costs        # Cost breakdown by token type
cburn status       # Claude.ai subscription status

CLI Commands

Command	Description
cburn	Usage summary (default)
cburn summary	Detailed usage summary with costs
cburn costs	Cost breakdown by token type and model
cburn daily	Daily usage table
cburn hourly	Activity by hour of day
cburn sessions	Session list with details
cburn models	Model usage breakdown
cburn projects	Project usage ranking
cburn status	Claude.ai subscription status and rate limits
cburn config	Show current configuration
cburn setup	Interactive first-time setup wizard
cburn tui	Interactive dashboard

Global Flags

-n, --days INT        Time window in days (default: 30)
-p, --project STRING  Filter to project (substring match)
-m, --model STRING    Filter to model (substring match)
-d, --data-dir PATH   Claude data directory (default: ~/.claude)
-q, --quiet           Suppress progress output
    --no-cache        Skip SQLite cache, reparse everything
    --no-subagents    Exclude subagent sessions

Examples:

cburn -n 7                      # Last 7 days
cburn costs -p myproject        # Costs for a specific project
cburn sessions -m opus          # Sessions using Opus models
cburn daily --no-subagents      # Exclude spawned agents

TUI Dashboard

Launch with cburn tui. Navigate with keyboard:

Key	Action
o / c / s / b / x	Jump to Overview / Costs / Sessions / Breakdown / Settings
<- / ->	Previous / Next tab
j / k	Navigate lists
J / K	Scroll detail pane
Ctrl+d / Ctrl+u	Scroll half-page
Enter / f	Expand session full-screen
Esc	Back to split view
r	Refresh data
R	Toggle auto-refresh
?	Help overlay
q	Quit

Tabs

• Overview - Summary stats, daily activity chart, live hourly/minute charts
• Costs - Cost breakdown by token type and model, cache savings
• Sessions - Browseable session list with detail pane
• Breakdown - Model and project rankings
• Settings - Configuration management

Themes

Four color themes are available:

• flexoki-dark (default) - Warm earth tones
• catppuccin-mocha - Pastel colors
• tokyo-night - Cool blue/purple
• terminal - ANSI 16 colors only

Change via cburn setup or edit ~/.config/cburn/config.toml.

Configuration

Config file: ~/.config/cburn/config.toml

[general]
default_days = 30
include_subagents = true

[claude_ai]
session_key = "sk-ant-sid..."    # For subscription/rate limit data

[admin_api]
api_key = "sk-ant-admin-..."     # For billing API (optional)

[appearance]
theme = "flexoki-dark"

[budget]
monthly_usd = 100                 # Optional spending cap

[tui]
auto_refresh = true
refresh_interval_sec = 30

Environment Variables

Variable	Description
CLAUDE_SESSION_KEY	Claude.ai session key (overrides config)
ANTHROPIC_ADMIN_KEY	Admin API key (overrides config)

Claude.ai Session Key

The session key enables:

• Real-time rate limit monitoring (5-hour and 7-day windows)
• Overage spend tracking
• Organization info

To get your session key:

1. Open claude.ai in your browser
2. DevTools (F12) > Application > Cookies > claude.ai
3. Copy the sessionKey value (starts with sk-ant-sid...)

Caching

Session data is cached in SQLite at ~/.cache/cburn/metrics_v2.db. The cache uses mtime-based diffing - unchanged files are not reparsed.

Force a full reparse with --no-cache.

Development

make build       # Build ./cburn binary
make install     # Install to ~/go/bin
make lint        # Run golangci-lint
make test        # Run unit tests
make test-race   # Tests with race detector
make bench       # Pipeline benchmarks
make fuzz        # Fuzz the JSONL parser (30s default)
make clean       # Remove binary and test cache

Architecture

~/.claude/projects/**/*.jsonl
    -> source.ScanDir() + source.ParseFile()  (parallel parsing)
    -> store.Cache (SQLite, mtime-based incremental)
    -> pipeline.Aggregate*() functions
    -> CLI renderers (cmd/) or TUI tabs (internal/tui/)

Package	Role
cmd/	Cobra CLI commands
internal/source	File discovery and JSONL parsing
internal/pipeline	ETL orchestration and aggregation
internal/store	SQLite cache layer
internal/model	Domain types
internal/config	TOML config and pricing tables
internal/cli	Terminal formatting
internal/claudeai	Claude.ai API client
internal/tui	Bubble Tea dashboard
internal/tui/components	Reusable TUI components
internal/tui/theme	Color schemes

License

MIT