§ 001 / AFK WRAPPED
Your agents had a year.
See what your AI coding agents did while you were away: the hours, your top tools, your agent persona, and a nine-slide story to post. One command. It reads your local Claude Code logs, and nothing leaves your machine unless you ask for a rank.
Want a rank? Add --rank, and --handle you to put your name on the board.
No account. No API key. No telemetry. Is it safe to run?
§ 002 / THE BOARD · 2026
Hours of agent work while away
- 1anonymousNIGHT SHIFT MANAGER264 h
§ 003 / SIX PERSONAS
Which one are you?
THE NIGHT SHIFT MANAGER
You go to bed. They clock in.
THE SWARM LORD
One agent is a tool. Twenty at once is a workforce.
THE ABSENTEE LANDLORD
You set the rules and left the building.
THE BASH GOBLIN
If it can't run in a terminal, it can't run.
THE MICROMANAGER
You check in. A lot.
THE PAIR PROGRAMMER
Side by side, prompt by prompt.
The command tells you which. Tag the friend who is a Bash Goblin.
§ 004 / SAFETY
Is it safe to run?
- It runs on your machine and only reads your Claude Code logs in ~/.claude/projects.
- It never changes your logs. It writes the PNGs to one output folder. With --rank it also keeps a random key in ~/.afk-wrapped.
- It sends nothing unless you add --rank. No account, no API key, no telemetry.
- No install scripts and one dependency, the image renderer. About a thousand lines of plain JavaScript.
Read all the code on GitHub (MIT), or the exact files on npm, before you run it.
What --rank sends
- Totals only: hours, counts, your top 5 tool names and file types, and the 24-hour shape of your work.
- Never code, prompts, paths or project names. MCP tools show only as "MCP".
- An optional handle. The CLI prints exactly what it sends before it sends it.
- Numbers are self-reported, so the board is for fun, not proof.
Not affiliated with Anthropic. Works with Claude Code logs.