Analytics

Running several coding agents leaves a trail in half a dozen formats — JSONL token logs, per-CLI session stores, prompt history, git — and none of them answer the questions you actually have: how much did the agents do this week, which one burned the tokens, were they really running in parallel, and what would it have cost at list prices. Analytics reads that trail locally and lays it out across four tabs. **Overview** gives you agent sessions against the previous period, prompts sent, agent time estimated from active 5-minute windows, tokens with the cache share, estimated spend, a weekday-by-hour punchcard, top projects, what you shipped in git and your longest sessions. **Agents** breaks tokens down per CLI into cache read, cache write, output and input, and lists the models you run with a warning on any model that has no known price. **Sessions** draws the week as a timeline with a parallel-agents chart and your orchestration runs passed and failed, and **Projects** puts prompts, sessions, commits and lines side by side. Scope any of it to 24h, 7d, 30d, 90d or All, and to one project or one agent. Open it with ⌘⇧A, from Navigate → Analytics…, or from the Command Palette.