activate

name	activate
description	Activate available managed agent memory
argument-hint	[task or retrieval cue]
user-invocable	true

Call the activate MCP tool once with a natural-language query describing the current task or retrieval cue. Use $ARGUMENTS when provided; otherwise derive the query from the current user task.

Apply the returned fragments directly. Pass a prior next_state only when its earlier fragments are still present in the current model context; omit state when starting fresh or after context compaction.

$ARGUMENTS