Token usage statistics for CLI coding agents — Claude Code, Codex CLI, Kimi Code, and Gemini.
The name is a double metaphor: tokens are the toll you pay to use AI coding agents, and heavy usage takes a toll.
- Tracks token usage from Claude Code (
~/.claude/projects/), Codex CLI (~/.codex/sessions/), Kimi Code (~/.kimi/sessions/), and Gemini (~/.gemini/tmp/) - Shows input, output, cached tokens, cache hit rate, and estimated USD cost
- Uses compact
k/m/btoken units by default, with--detailfor raw counts - Per-model breakdown across all sessions
- Filter by today, last N days, or all time
- View per-tool or combined stats
- List all supported model prices with
--list-prices
| Tool | Supported versions | Log format |
|---|---|---|
| Claude Code | All versions (JSONL logs present since launch) | message.usage entries in ~/.claude/projects/**/*.jsonl |
| Codex CLI | ≥ 0.1 (rollout log format) | token_count events in ~/.codex/sessions/YYYY/MM/DD/rollout-*.jsonl |
| Kimi Code | All versions | StatusUpdate token_usage events in ~/.kimi/sessions/**/<session>/wire.jsonl |
| Gemini | ≥ 0.5 (antigravity log format) | tokens entries in ~/.gemini/tmp/**/chats/*.json |
Downloads the latest GitHub Release for your platform. If a prebuilt binary is not available, the installer falls back to cargo install toll.
curl -fsSL https://raw.githubusercontent.com/Fullstop000/toll/refs/heads/master/install.sh | shInstall to a custom directory:
curl -fsSL https://raw.githubusercontent.com/Fullstop000/toll/refs/heads/master/install.sh | TOLL_INSTALL_DIR="$HOME/.local/bin" shRequires Rust.
cargo install tollgit clone https://github.com/Fullstop000/toll.git
cd toll
cargo install --path .Token usage statistics for Claude Code, Codex CLI, Kimi Code, and Gemini
Usage: toll [OPTIONS]
Options:
-v, --version Show version information and exit
--today Show today's usage only
--days <N> Show last N days
--claude Show Claude stats only
--codex Show Codex stats only
--kimi Show Kimi Code stats only
--gemini Show Gemini stats only
--json Emit JSON to stdout
--csv Emit CSV to stdout
--list-prices List all supported models and their prices, then exit
--detail Show full token counts instead of compact b/m/k units
--by-day Show usage aggregated by day
-h, --help Print help
Examples:
toll # all-time stats
toll --today # today only
toll --days 7 # last 7 days
toll --by-day --days 7 # daily summary table
toll --claude # Claude only
toll --codex # Codex only
toll --kimi # Kimi Code only
toll --gemini # Gemini only
toll --detail # full token counts
toll --json # machine-readable JSON
toll --csv # terminal CSV export
Token usage — all time
Collected: 2026-03-13 20:20:38 +08:00
Sessions Input Cached Hit Rate Net Input Output Total Cost
════════════════════════════════════════════════════════════════════════════════════
Claude Code 38 316.1m 304.2m 96.2% 11.9m 1.3m 317.5m $168.59
Codex 21 148.9m 137.4m 92.3% 11.5m 777.8k 149.7m $74.70
Kimi Code 15 34.6m 32.4m 93.6% 2.2m 206.2k 34.8m $6.70
Gemini 10 28.5m 25.8m 90.5% 2.7m 185.5k 28.7m $2.15
────────────────────────────────────────────────────────────────────────────────────
Combined 114 528.1m 499.8m 94.6% 28.3m 2.5m 530.6m $252.14
────────────────────────────────────────────────────────────────────────────────────
By model:
─────────────────────────────────────────────────────────────────────────────────
Model Tokens Output Cost
─────────────────────────────────────────────────────────────────────────────────
claude-haiku-4-5-20251001 9.4m 40.5k $2.71
claude-opus-4-6 49.7m 317.2k $45.17
claude-sonnet-4-6 258.3m 988.2k $120.72
gemini-3-flash-preview 28.7m 185.5k $2.15
gpt-5.4 149.7m 777.8k $74.70
kimi-for-coding 34.8m 206.2k $6.70
─────────────────────────────────────────────────────────────────────────────────
Scanned 114 session(s) in 1.42s
| Tool | Log path | Token field |
|---|---|---|
| Claude Code | ~/.claude/projects/**/*.jsonl |
message.usage per API call — sums input_tokens, cache_creation_input_tokens, cache_read_input_tokens, output_tokens |
| Codex CLI | ~/.codex/sessions/YYYY/MM/DD/rollout-*.jsonl |
last token_count event per session (cumulative totals) |
| Kimi Code | ~/.kimi/sessions/**/<session>/wire.jsonl |
StatusUpdate events — sums input_other, input_cache_read, input_cache_creation, output per API call |
| Gemini | ~/.gemini/tmp/**/chats/*.json |
tokens per agent message — sums input, cached, output per call; totals can differ from Gemini CLI /stats model because Gemini CLI also counts internal utility-model requests, while toll reads only persisted chat turns |
MIT