v0.7.0 — Usage, Coverage & Token Hygiene

See your usage, spend less. v0.7.0 adds three pillars for getting more Claude Code for less — all local-first, all keyed off your own claude CLI.

📊 Measure — usage ledger

Every claude call Adeptly makes (recipe, crew, feature-refresh) now records its real token/cost receipt (--output-format json) to .adeptly/usage.jsonl. The new Usage panel shows spend, cache-hit %, and cost by source & model, over 7d / 30d / all-time. No telemetry — the ledger never leaves your machine.

💰 Estimate — cost up front

The recipe's expected turns + cost are surfaced, and the Crew shows estimate vs. actual after a live run.

🌿 Optimize — Token Hygiene (Savings tab)

An efficiency score for the plan against the money-saving Claude habits — Plan Mode first, Explore subagent for search, /clear, cache-friendly ordering, Haiku for mechanical work, background agents, fewer permission prompts. Each missing habit is one click to bake into the plan.

🎯 Coverage

Every plan scored on how many of Claude Code's 30+ features it actually uses, with the high-value gaps (and why) — one click to add each. The core pitch, made measurable.

🔄 Refresh the catalogue from your CLI

The built-in feature list is a snapshot that goes stale. Hit Refresh and Adeptly asks your own (self-updating) claude CLI what's new, validates + de-dupes it, and lets you accept additions into .adeptly/features.json. No backend, no doc-scraping.

---

Install / upgrade:

npm install -g adeptly@0.7.0   # or: npx adeptly

Full parity with npm adeptly@0.7.0.