TokOp — Token Optimizer for Claude Code

Route every AI task to the cheapest model that can actually finish it — and never let use-it-or-lose-it plan quota reset unused. THE PROBLEM (measured, not assumed): 241.6M tokens in 7 days, ~$126 at API list price. 200.4M ran on flagship-tier models; zero on the cheap tier — even though much of it was routine formatting, edits, and file wrangling a cheap model would have finished for a fraction of the cost. WHAT IT DOES: model registry with capability + strengths data for Claude, OpenAI and Gemini models; usage metering that reads your real local Claude Code logs and shows exactly where tokens go; a smart router that grades each task (trivial to hard/creative) and sends it to the cheapest capable model, escalating only if the cheap one fails; a quota scheduler that watches your plan reset window and drains pre-approved backlog before unused capacity expires. INSIDE: 9 Python modules, full README with architecture docs, model capability registry, AI strengths guide, efficiency notes, Windows install script. REQUIRES: Python 3.10+, Claude Code. Works with API billing or plan quota. This is the working-software counterpart to the Token Optimizer Kit ($29 policy guide) — the kit teaches the method, TokOp runs it.