claudetop

---

14:32  my-project/src/app  Fable 5.1  20m 0s  +256/-43  #auth-refactor
152.3K in / 45.2K out  ████░░░░░░ 38%  $3.47  $5.10/hr  ~$174/mo
cache: 66%  efficiency: $0.012/line  opus:~$3.20  sonnet:~$0.88  haiku:~$0.23
plan: 5h ████████░░ 82% ↻2h18m  7d ██████░░░░ 64% ↻4d1h  Fable ████████░░ 81% ↻4d1h  credits $0/$2000
in:80% out:20% (fresh:15% cwrite:7% cread:76%)
$5 MARK  |  main*  |  ♫ Artist - Song  |  PROJ-123  |  CI ✓

The problem

Claude Code doesn't show you what you're spending. You finish a session, check your billing dashboard, and discover a $65 charge for what felt like 30 minutes of work. You have no idea which session caused it, which model was wasteful, or whether your cache was even working.

I built claudetop after noticing my model estimate showed $10 but the actual bill was $65. Turns out, compaction was hiding 80% of my token usage. The cost was real — the visibility wasn't.

Install

Clone and install (recommended)

git clone https://github.com/liorwn/claudetop.git
cd claudetop && ./install.sh

One-liner

curl -fsSL https://raw.githubusercontent.com/liorwn/claudetop/main/install.sh | bash

As a Claude Code plugin

claude plugin marketplace add liorwn/claudetop
claude plugin install claudetop

This gives you the SessionEnd hook + all slash commands (/claudetop:stats, /claudetop:dashboard, /claudetop:branch, /claudetop:export, /claudetop:pricing) automatically.

Then restart Claude Code.

What you see

Before claudetop

>

A blank prompt. No context. No cost. No idea.

After claudetop

14:32  my-project/src/app  Fable 5.1  20m 0s  +256/-43  #auth-refactor
152.3K in / 45.2K out  ████░░░░░░ 38%  $3.47  $5.10/hr  ~$174/mo
cache: 66%  efficiency: $0.012/line  opus:~$3.20  sonnet:~$0.88  haiku:~$0.23
plan: 5h ████████░░ 82% ↻2h18m  7d ██████░░░░ 64% ↻4d1h  Fable ████████░░ 81% ↻4d1h  credits $0/$2000
$5 MARK  |  TRY /fast  |  main*  |  CI ✓  |  ♫ Bonobo - Kerala

Every response, you see:

• What project you're in and how deep
• What model is running and for how long
• What it costs right now, per hour, and projected monthly
• How much plan you have left — 5-hour, 7-day, and per-model (Fable/Opus/Sonnet) windows with reset countdowns
• How efficient your cache is (are you wasting tokens?)
• What it would cost on a different model (should you switch?)
• Smart alerts when something is wrong

Features

Plan usage windows (subscribers)

If you're on a Claude Pro/Max/Team plan, the plan: line shows every rate-limit window Anthropic enforces on you — as a bar, % used, and time until it resets:

plan: 5h ████████░░ 82% ↻2h18m  7d ██████░░░░ 64% ↻4d1h  Fable ████████░░ 81% ↻4d1h  credits $0/$2000

Window	What it is
5h	Current session window (rolls every 5 hours)
7d	Current week, all models
Fable / Opus / Sonnet	Current week, that model only — appears whenever your plan has a per-model cap
credits	Extra-usage credits spent this month / your monthly cap (only if extra usage is enabled)

Bars go yellow at 50% and red at 80%. At 90% you get a Fable LIMIT 93% ↻4d1h alert so you can switch models or wrap up before Claude Code stops you. The minimal theme shows the same numbers inline (5h:82% 7d:64% Fable:81%); compact shows only the tightest window.

The 5h and 7d numbers come straight from Claude Code (the status line JSON carries them after the first API response). Per-model windows and credits come from the same usage endpoint the /usage screen uses, fetched by claudetop-usage in the background at most once a minute and cached in ~/.claude/claudetop-usage.json — the status line itself never waits on the network. It reads the OAuth credential Claude Code already stored (macOS Keychain, or ~/.claude/.credentials.json on Linux) and never refreshes or modifies it. API-key, Bedrock, and Vertex sessions have no plan windows, so the line simply doesn't appear.

claudetop-usage --show       # every window as a 20-char bar with % used, % left, reset time
claudetop-usage --refresh    # force a refetch
export CLAUDETOP_USAGE=off   # hide the plan line
export CLAUDETOP_USAGE_TTL=120   # refresh every 2 minutes instead of 1

Or in Claude Code: /claudetop:usage.

Real-time cost tracking

Your actual session cost (green), burn rate per hour, and monthly forecast extrapolated from your history. No more billing surprises.

Model cost comparison

See what your session would cost on Opus, Sonnet, or Haiku — with cache-aware pricing that accounts for your actual cache hit ratio. The current model is bolded so you can instantly compare.

Pricing updates automatically from the pricing.json in this repo — when Anthropic changes prices, claudetop stays current.

Cache efficiency

Your cache hit ratio tells you if you're being efficient. Green (≥60%) means most of your input tokens are being reused. Red (<30%) means something is forcing full re-reads — maybe compaction, maybe a model switch.

Smart alerts

Only appear when something needs your attention:

Alert	What happened	What to do
$5 MARK / $10 / $25	Cost milestone crossed	Gut-check: am I getting value?
OVER BUDGET	Daily budget exceeded	Wrap up or switch models
CONSIDER FRESH SESSION	>2hrs + >60% context	Start fresh — diminishing returns
LOW CACHE	<20% cache after 5min	Context was reset, tokens being re-read
BURN RATE	>$15/hr velocity	Runaway subagents or tight loops
SPINNING?	>$1 spent, zero code output	Stuck in a research loop
TRY /fast	>$0.05/line on Opus	This task doesn't need the biggest model
COMPACT SOON	Context window >80% full	Auto-compaction is imminent
Fable LIMIT 93%	A plan window (5h / 7d / per-model) is ≥90% used	Switch models or wrap up; shows when it resets

Session history & analytics

Every session is automatically logged. See where your money goes:

claudetop-stats              # Today's summary
claudetop-stats week         # This week
claudetop-stats month        # This month
claudetop-stats all          # All time
claudetop-stats tag auth     # Filter by tag

claudetop-stats  This Week
──────────────────────────────────────────────────────

Summary
  Sessions:     12
  Total cost:   $47.30
  Avg / session:  $3.94
  Daily avg:      $9.46

Cost by model
  claude-opus-4-6:       $38.20
  claude-sonnet-4-6:     $9.10

Top projects by cost
  rri-os                 $22.50  (4 sessions)
  pistol-claw            $14.80  (5 sessions)
  the-table              $10.00  (3 sessions)

Deep analytics engine

The analytics engine scans all Claude Code JSONL session files — not just hook-recorded ones. It captures your complete history including sessions from before claudetop was installed, subagent costs, and per-turn token breakdowns.

claudetop-engine scan       # Scan all JSONL files → SQLite
claudetop-engine today      # Today (with per-turn tool breakdown)
claudetop-engine stats      # All-time (subagents, top tools, projects)
claudetop-engine dashboard  # Live web dashboard at localhost:8080

The live dashboard features an activity heatmap, cost-by-model charts, project and tool breakdowns, subagent cost attribution, and a sortable session table — all with model filtering and time range selection. Auto-refreshes every 30 seconds.

Zero dependencies — uses only Python 3 standard library (sqlite3, http.server, json).

Session tagging

Track costs per feature, bug, or initiative:

export CLAUDETOP_TAG=auth-refactor
# ... work on auth ...
claudetop-stats tag auth-refactor
# Total cost: $12.40 across 3 sessions

Daily budget

export CLAUDETOP_DAILY_BUDGET=50

Shows budget: $12 left at 80% → OVER BUDGET ($52/$50) when exceeded.

Themes

export CLAUDETOP_THEME=full      # Default: 3-5 lines
export CLAUDETOP_THEME=minimal   # 2 lines
export CLAUDETOP_THEME=compact   # 1 line

iTerm2 integration

Push claudetop data into iTerm2's chrome — tab titles, status bar, and badge watermark:

export CLAUDETOP_ITERM=all           # Enable everything
export CLAUDETOP_ITERM=title         # Tab/window title only
export CLAUDETOP_ITERM=badge         # Watermark overlay only
export CLAUDETOP_ITERM=statusbar     # User variables for status bar
export CLAUDETOP_ITERM=bgcolor       # Background color tint by state
export CLAUDETOP_ITERM=title,badge   # Combine any options

Tab title — Shows project | $4.21 | Fable 5.1 | ctx:38% | Fable 81% in your iTerm2 tab (the last item is your tightest plan window). Zero configuration.

Badge — Faint watermark in the terminal background with cost, model, and context at a glance. Great for keeping cost visible while scrolling through output.

Background color — Subtly tints the terminal background based on session state:

Tint	Meaning
Green	Healthy session (low context, under $5) or session ended (idle/waiting)
Amber	Caution — cost milestone, compact soon, low cache, spinning
Red	Alert — over budget, burn rate spike, $25 mark
Default	Normal — no special state, uses your profile's background

When a session ends, the background stays green so you can see at a glance which terminals are idle vs active. Resets to default when the next session starts clean.

Status bar — Sets iTerm2 user-defined variables that you can display in the status bar (top or bottom of terminal). Configure in iTerm2: Preferences > Profiles > Session > Status Bar > add "Interpolated String" components:

Variable	Content	Example
\(user.claudetop_cost)	Session cost	$4.21
\(user.claudetop_model)	Current model	Opus 4.6
\(user.claudetop_ctx)	Context usage	38%
\(user.claudetop_project)	Project name	my-project
\(user.claudetop_duration)	Session time	20m 0s
\(user.claudetop_cache)	Cache hit ratio	66%
\(user.claudetop_velocity)	Burn rate	$5.10/hr
\(user.claudetop_tokens_in)	Input tokens	152.3K
\(user.claudetop_tokens_out)	Output tokens	45.2K
\(user.claudetop_lines)	Lines changed	+256/-43
\(user.claudetop_tag)	Session tag	#auth-refactor
\(user.claudetop_usage)	All plan windows	5h:82% 7d:64% Fable:81%
\(user.claudetop_usage_top)	Tightest plan window	Fable 81%

No-op on non-iTerm2 terminals — escape sequences are silently ignored.

Context composition

See what's eating your context window:

in:80% out:20% (fresh:15% cwrite:7% cread:76%)

High cread = cache is working well. High fresh = re-reading files every turn.

Plugins

Drop any executable script into ~/.claude/claudetop.d/ — it becomes part of your status line.

Included (enabled by default):

Plugin	What it shows
git-branch.sh	main* (branch + dirty indicator)

Example plugins (copy to enable):