CC One 2.3.0

Your AI CLI usage, owned by you. cc one reads the session logs your AI CLIs already write and turns them into tokens, cost, cache efficiency, and trends -- a local-first dashboard with optional multi-device sync through a GitHub repo you control.

Why

Every session with Claude Code, Codex, Gemini CLI, Grok CLI, or OpenCode writes a log to disk -- tokens, cache hits, spend -- all in plain text, unread. cc one reads those logs and turns them into a clear picture: what you spent, what you received, and where your tokens went. It also manages provider switching, writing the exact config file each CLI reads.
Two principles govern the design:
- Local-first. The full dashboard works with zero network -- the logs already on your disk are all it needs.
- Read-only toward your logs. cc one never modifies session logs or tool behavior. The sole write is provider switching: always your explicit action, always backed up first.
Multi-device sync is an opt-in layer, never a precondition.

Features

Providers
- One config hub for all five CLIs -- switching a provider writes the real config file that CLI reads, merging controlled fields only, with the previous file backed up first.
- 59 built-in presets -- select one, enter your API key, and switch. Import from CC-Switch, local config files, or a CC One backup; export the whole list as JSON anytime.
- Model role mapping -- five roles (Sonnet / Opus / Fable / Haiku / Subagent), each with its own model and a 1M-context toggle; one click fetches the vendor's model list.
- Structure syncs, keys never -- the provider list syncs per device; API keys are stripped from anything that leaves your machine.

Dashboard
- Four-bucket token economics -- input / output / cache creation / cache read, normalized across CLIs into one model that matches your bill.
- Cache-hit rate, model distribution, requests & cost -- models ranked by tokens, each with its own cache-hit rate; totals frozen at collection time.
- Trends & request log -- multi-metric trend chart with today-vs-yesterday delta; per-call log with token breakdown, cost, and full details on click.

Sessions
- Browsable history -- every conversation grouped by project, full-text search, filters by time / source / model / device.
- Complete transcripts, instantly -- stored locally at collect time, rendered as markdown with syntax highlighting; in-session search with hit-highlighting and jump.
- Local & Favorites -- private local groups that never leave this machine; favorited sessions sync across devices with source-device badges. Only favorited sessions ever leave your machine.

Sync (optional)
- Your own repo, plain text -- usage projected into human-readable, per-device, per-day JSONL in a GitHub repo you control; no third-party server in the middle.
- Conflict-free by construction -- device-isolated writes, deterministic artifacts, self-healing rebases.
- System-proxy aware; dashboard, glance card, and tucked bar can be scoped to a single device.

Library
- Drag-to-relay upload -- drop files or directories into the window; they travel through your sync repo into that device's subtree.
- In-app preview & manual export -- images, text/JSON, sandboxed rendering for the rest; export to a path you choose.

Cost & pricing
- Editable per-model pricing, one-click LiteLLM cost-map fetch, rebill for records collected without a price, JSON import/export.

Experience
- Glance mode -- a tucked mini-bar or a floating card mirroring the dashboard; each shape retains its placement.
- Multi-skin theming, tray-resident background collection (5s-60s), signed auto-updates, UI in English / ???? / ???.

What stays local, what syncs

| | Standalone | Synced (repo bound) |
| ------------------------------------------------ | ---------- | -------------------------------------------- |
| Usage, cost, trends, request log | Local only | Syncs across devices |
| Session transcripts & favorites | Local only | Favorited sessions sync; the rest stay local |
| Provider structure | Local only |...

License: MIT