Learn kal your way
Everything here is short, shown with real terminal output, and drawn wherever a picture beats a paragraph. Pick a starting point.
The 5-minute tour
Eight steps from empty folder to a history that explains itself: intent-first snapshots, AI attribution, blame, undo, verify, ask, git sync.
Start the tour →Command reference
All 40+ commands in six themed groups, each with a plain-language explanation and the exact output it prints. Live filter included.
Browse commands →How kal thinks
Five ideas, five diagrams: the intent timeline, immutable history with revisable notes, the undo log, the trust ladder, and the git bridge.
Read the concepts →Coming from git
Your muscle memory mostly works. The translation table, what improves, and an honest answer to “where did the staging area go?”
See the mapping →AI agents & platforms
MCP in one command, the three-variable attribution contract, Claude Code hooks, AGENTS.md generation, model scorecards, AZMX Model wiring.
Integrate an agent →Install kal
brew, deb, rpm, nix, Windows installer, portable exe, pipx, or a universal zipapp — every artifact CI-tested on its own platform.
Get kal →The docs travel with the binary
Everything on this site is in the CLI: bare kal prints the grouped
overview, kal help <command> details any command, and kal help tour
runs the tour in your terminal.
$ kal help snap usage: kal snap [-h] [-m MESSAGE] [-i INTENT] [--agent AGENT] [--model MODEL] [--amend] [--json] options: -h, --help show this help message and exit -m, --message MESSAGE -i, --intent INTENT why this change exists — the question future you (or an agent) will ask --agent AGENT record an AI agent as the author --model MODEL model id behind --agent --amend replace the tip snapshot (old one is kept; undoable) --json