01 / Aggregate locally
Start with totals you can inspect.
The open-source CLI uses ccusage to read supported coding-agent logs on your machine, including Claude Code, Codex, Gemini CLI, and Qwen. It groups activity into daily totals. Before a sync, you can run a dry run to collect usage without submitting it. Straude is designed to measure the shape of your practice without needing the substance of the work that produced it.
That makes setup intentionally small: one command after a session, no project instrumentation, no repository access, and no browser extension watching how you work.
02 / Log the session
Keep the metrics that help you train.
A Straude day can include token volume, estimated cost, model names, and session count. Those aggregates become a consistent record you can review across days instead of a pile of disconnected provider logs. Streaks make consistency visible, while profiles and activity posts give you a durable record of showing up.
Cost figures are estimates for reflection and comparison. Provider invoices remain the authority for billing, and community statistics represent participating Straude users rather than every developer.
Over time, the log helps answer practical questions: whether your usage is rising, which models shape your workload, how often you return to the tools, and which weeks produced your strongest pace. The point is a clearer feedback loop, not a single score that decides whether a session was worthwhile.
03 / Protect the boundary
Leave prompts and code where they belong.
Prompts, conversations, source code, project names, and file contents are outside Straude's collection boundary. They remain on your machine. Straude does not analyze code quality, reconstruct what you asked an agent, or monitor employees. It receives the aggregate usage totals you choose to send.
Read the privacy policy for the full data boundary, or explore the CLI reference before your first sync.