Everything You Need
To Control AI Spend

15 real, shipped features — zero-config session tracking, an optimization engine, and privacy-first observability, all running on your machine.

Zero-config tracking across 7 tools: Claude Code, Cursor, Aider, GitHub Copilot, Windsurf, Cline, and OpenAI Codex CLI.

A tree view of every subagent Claude Code spawns — cost, tools used, and duration per agent.

20+ rules analyze your usage patterns and surface specific, actionable savings.

Opt-in lexical/statistical scoring flags when a project’s responses start diverging from their own baseline.

Statistically honest comparison (two-sample z-tests) between models, projects, or time windows — built from data you already have.

Step-by-step tool-call reconstruction for any single request, right in Request Detail.

Opt-in, off by default. Detects email, phone, SSN, credit card, and API keys in proxied traffic.

Per-provider, per-model daily/weekly/monthly limits with a hard kill switch and pre-request cost estimation.

OS-level connection detection attributes cost to the app driving it — zero per-app config needed.

Opt-in, bring your own key: a Claude-generated spend summary built only from aggregated metadata, never your prompts.

A real health score showing identified savings as a multiple of what the Pro plan itself costs.

Rolling rate-limit windows and a 24h activity heatmap so you can route around peak usage.

Monthly and yearly spend reports with shareable cards, built from four real insight algorithms.

Webhook alert rules for budget thresholds and anomaly spikes — Slack, Discord, or any HTTP endpoint.

Session data is parsed from files your tools already write. Nothing leaves your machine unless you opt in.