LIVE v1.2 · LiteLLM pricing · 15 agents · MIT

Guessing every
token your agents
actually spend.

TokenMeter is a local-first usage meter for Claude Code, Codex, and 13 other AI coding agents. Real-time hooks. LiteLLM-accurate pricing. A live dashboard — entirely on your machine.

Install TokenMeter →
$ brew install --cask tt-a1i/tap/tm

Live · what TokenMeter sees right now

data refreshed 4s ago · sqlite read
SPEND · MAY 2026
$284.62
↑ 18.2% vs Apr · 71% of $400 budget
$0$400 budget
7 PROJECTS · 12 MODELS · 187 SESSIONS
5-HOUR BLOCK
82%
14:00 → 19:00
BURN 340 tok/min · PROJ 18:42
TOKENS · 14 DAYS
12.4M
↓ 3.1% vs prev · avg 412k / day
LIVE EVENT STREAM · tm watch SOCKET ATTACHED
COST BY MODEL · 30D USD
opus-4$192.41
gpt-5$71.20
sonnet-4$31.84
codex-5$21.20
haiku-4$8.71
gemini-2.5$4.42
12 MODELS ACTIVE · LITELLM v1.74 · 24H TTL
USAGE HEATMAP · 280 HOURS UTC · GREEN = HOT
14 DAYS · MAY 12 → MAY 25 low → high
SUPPORTED SOURCES v1.2
15 agents
Claude Code, Codex, OpenCode, Amp, Gemini CLI, Copilot, Goose, Codebuff, Hermes, Kilo, Kimi, OpenClaw, pi-agent, Droid & Qwen.
5H BLOCKS · TODAY 5 SLOTS
00:0005:0010:0015:0020:00
CURRENT 14:00 → 19:00 · 82% LIMIT · PROJ 18:42
BUDGET · MAY $400.00
Spent$284.62
Remaining$115.38
Pace+12%
Alert at90%
TOOL ERRORS · 7D tm analyze
Bash"command not found"42
Edit"string not found"18
Read"file does not exist"9
Greptimeout · large repo4
$  tm daily --compact ZSH · ~/code
DAEMON HEALTH ● ONLINE
Uptime128h 42m
Socket/tmp/tokenmeter.sock
Hooks attached8 / 8
Hook latency~14ms p95
DB size42.1 MB
WEBHOOKS
3
SLACK · DISCORD · CUSTOM
TELEMETRY SENT
0 bytes
Nothing leaves your machine. MIT licensed, ~/.tokenmeter local-only.
15 AGENTS · 1 TIMELINE AUTO-DISCOVERED

Start metering in
under a minute.

One binary, zero accounts, zero telemetry. Everything lives in ~/.tokenmeter — yours alone.

Install TokenMeter → View on GitHub
$ brew install --cask tt-a1i/tap/tm
copied