Introduction

SessClone collects Claude Code usage from every machine your team works on and puts it in one place, so an Org can see what it is spending — per Member, per Project, per Device — and, when a Member opts in, download the raw session logs to learn from them.

What is collected

A Claude Code plugin, the Collector, runs on each machine and reports usage only: token counts, models, timings, and the Session, Project and Device they belong to — plus, per Turn, the working directory it ran in, the git branch, and the git remote normalised into a Project key. Never a prompt, never a file's contents.

Transcripts — the raw session logs, which do contain both — are a separate upload that stays off until a Member turns it on.

Reading Costs

• Over time draws a bar for every day of the period, including empty days. A bar, a day row or a model row opens a column: a day shows its tokens by class and by model, then the Sessions that ran that day. A model shows its tokens by class (input, output, cache read, cache write), each with its cost, then the Sessions that used it.
• People, Projects and Devices open the same column for a row: its token table, a compact table by model, then its Sessions. Every column lists Sessions, not Turns, each with its name, start time, cost and tokens, and links to the Session.
• Counts outside a Session's own page are sessions, not turns. A Session of many Turns counts once.
• Failures counts failed Sessions you have not marked viewed.

Where to start