AI agents

Every session gets a private context file:

$ katu join api
$ echo "$KATU_AI_CONTEXT"
/home/you/.katu/ai-contexts/api.json

External AI CLIs read and write that file; switching sessions switches the file. To hand your assistant the current terminal output:

katu capture | claude "explain this database lock error"

Managing agents

katu ai              # floating AI popup, or a dedicated ai-<dir> session
katu agents          # list agents: status, name, age, location
katu agents --pick   # fzf picker; enter jumps to the agent's pane

alt a toggles the AI popup — a floating pane over your work, tmux display-popup style. alt u opens the native agent picker with a live output preview, and alt p gives you a scratch-shell popup.

The status line marks any window hosting an agent with a robot icon colored by state — yellow waiting for your input, red busy, green idle — so a glance at the bar tells you which agent needs attention. When an agent starts waiting, Katu also fires an OSC 9 terminal notification ([ai] notify).

Scrolling an agent

Full-screen agents like Claude Code run on the alternate screen and scroll their own transcript on the mouse wheel, so copy mode has nothing to browse. alt < / alt > scroll the focused pane by sending it wheel events — so you can page back through the conversation in the AI popup even without a mouse.