token-counter
by mkhaytman87
Coding Agents & IDEsBrief
Track and analyze OpenClaw token usage across main, cron, and sub-agent sessions with category, client, model.
Description
What This Skill Does
Parses local OpenClaw session transcripts, metadata, and cron definitions to produce token usage reports. Breaks down consumption by category, client, model, and tool across main, cron, and sub-agent sessions. Supports per-session drilldowns, time-range filters, and JSON export.
Reads directly from local transcript files rather than requiring a dashboard or external service, so reports reflect actual recorded usage with no data leaving the machine.
When to Use It
- Check where tokens went over the past week
- Get a per-tool breakdown before optimizing a cron job
- Export a 30-day usage snapshot for cost review
- Drill into a single session to find a token spike
- Save daily snapshots to track usage trends over time
Example Workflow
User asks: show me a breakdown of token usage by tool and category for the last 7 days
- 1
Run token-counter with --period 7d and --breakdown tools,category
- 2
Parser reads session transcripts from $OPENCLAW_DATA_DIR/agents/main/sessions/*.jsonl
- 3
Attribution logic splits assistant-message tokens across tool calls in each message
- 4
Aggregates counts by tool name and session category
Console report listing token totals per tool and per category for the 7-day window
Installation
clawhub install mkhaytman87/token-counterRun the CLI command, or paste the prompt into Claude Code, Cursor, Codex or whichever coding agent you use.