Dashboard
Spend, models, tokens, and latest sessions for Critique CLI and CritiqueCode — split so sidecar reviews never mix with the author agent.
Signed-in home is /dashboard. It is a usage ledger for the two live products, not the archived GitHub control room.
What it shows
| Column | Source |
|---|---|
| Spend | CLI cloud review fees + reported model cost; CritiqueCode Inference USD on a CritiqueCode device key |
| Top models | Models on those synced runs and Inference turns |
| Input / output | Tokens reported by finish runs and Inference |
| Latest sessions | Recent CLI finish runs and CritiqueCode Inference sessions |
Filter the session list with All, CLI, or CritiqueCode.
Critique CLI
critique login stores a machine key. Local and cloud critique finish summaries sync here. Source stays on the machine for local runs. Create or rotate keys under Account.
CritiqueCode
critique-code login or Login inside critique-code web issues a CritiqueCode device key. Author turns that hit Critique Inference appear on this dashboard. BYOK routes (OpenRouter, Anthropic, and the rest) bill the provider directly and do not show as Critique spend.
The local web UI also has its own session rail, token ring, and cost breakdown on the machine. The website dashboard is the account-level view of Inference-backed authoring.
Drive a running laptop from the browser at /code: Connect this PC, then critique-code connect CODE on the machine. The site relays prompts and run approvals; the kernel stays on the laptop.
Account vs Inference API
| Surface | Purpose |
|---|---|
| Dashboard | CLI + CritiqueCode spend and sessions |
| Account | Device keys, default review model, cloud env profiles |
| Inference API dashboard | Raw Inference API keys, caps, and request log |
Archived control room
The previous GitHub passports / repo-home / evidence-run workspace lives at /archive/dashboard. It is preserved, not promoted. Old /dashboard/passports, /dashboard/review-runs, and similar URLs redirect there.