Compute inference-only token costs from Claude Code JSONL logs.
npx skills add https://github.com/azigler/dotfiles --skill cost-tracking-azigler
Please help me install this Agent Skill. Skill: /cost-tracking Source: https://github.com/azigler/dotfiles/tree/main/agents/skills/cost-tracking Command: npx skills add https://github.com/azigler/dotfiles --skill cost-tracking-azigler
This Skill turns Claude Code session JSONL logs into a clear, project-level ledger of inference costs so you can understand where your token budget goes.
Set up cost tracking by copying the ledger template into .claude/plans/cost-tracking.md and then letting /offboard append new session rows automatically.
High-intent search queries and answers about installing and using this skill.
You can track Claude Code token usage costs by parsing session JSONL logs to compute an inference-only cost ledger. This sums orchestrator input and output tokens while filtering out subagent sidechain contributions to reflect main session work.
Anthropic API billing transparency is achieved by applying model-specific public pricing to filtered JSONL session logs, generating a per-project markdown ledger showing exact inference costs separated from cache token counters.
The cost ledger excludes cache token expenses from the headline cost. Cache tokens are tracked only as reference counters, while canonical cost is computed strictly from standard input and output token rates.
Automatic budgeting updates require copying the ledger template into .claude/plans/cost-tracking.md. Once the ledger exists, the /offboard command appends new session rows using the most recent JSONL log or a provided path.
Subagent sidechain token costs are excluded from the billing ledger to ensure totals reflect orchestrator work only. Filtering by type and isSidechain prevents sidechain contributions from inflating headline inference cost.
You can export Claude Code session cost tracking data as JSON. The ledger append logic supports emitting markdown rows, JSON, verbose output, or commit-lines for flexible integration with downstream budgeting and token-allocation analysis workflows.
aziglerCapture orchestrator-session state into a persistent handoff note and commit it to git.
aresbitAnalyze CodexBar cost logs to report per-model usage and costs.
brucebanner010198-commitsAppend project-close lessons to the cross-project LESSONS.md ledger.
netanel-abergelReport AI token usage and estimate USD costs from OpenClaw logs.
