proc: evaluator // net: none
What is your history worth?
Pick your Claude Code or Codex folder. The page counts your sessions right here and shows the card. Nothing is uploaded.
step 01 // pick a folder
- Claude Code
~/.claude/projects - Codex
~/.codex/sessions
These folders are hidden. On a Mac, press Cmd+Shift+. in the picker to show them, or Cmd+Shift+G and paste the path. On Windows they are under %USERPROFILE%. Pick your home folder's .claude to count everything.
Some browsers say "upload" here. It means this page may read the files; they stay on your computer. Switch off your network first if you like - the evaluator works offline.
your card appears here // nothing leaves this tab
How the estimate works
Sessions are sorted into tiers by length and tool use, and each tier gets the range buyers have reportedly paid. Short or chat-only sessions are not counted. Full method.
- Deep50+ turns, 1h+ active, tools used$40 - $200
- Working15+ turns, 15m+ active, tools used$5 - $40
What does the evaluator read?+
Timestamps, message roles and tool names from each session file. It counts turns, tool calls and active time, and names products only from a fixed list (GitHub, Supabase, Vercel and others). Message text and file contents are not kept.
How is active time measured?+
The time between consecutive entries in a session, skipping any gap over five minutes, so a session left open overnight counts only the minutes someone was working.
Which agents does it support?+
Claude Code (~/.claude/projects) and OpenAI Codex CLI (~/.codex/sessions) today. Other agents store history in different formats; see the agents section for where each one keeps it.