Basics

Core learning primitives and default patterns.

Example	Description
User Profile: Always Mode	ALWAYS mode extracts profile information automatically in parallel while the agent responds - no explicit tool calls needed.
User Profile: Agentic Mode	AGENTIC mode gives the agent explicit tools to update profile fields.
User Memory: Always Mode	ALWAYS mode extracts memories automatically in parallel while the agent responds - no explicit tool calls needed.
User Memory: Agentic Mode	AGENTIC mode gives the agent explicit tools to save and update memories.
Session Context: Summary Mode	Session Context tracks the current conversation’s state: - What’s been discussed - Key decisions made - Important context.
Session Context: Planning Mode	Session Context tracks the current conversation’s state: - What’s been discussed - Current goals and their status - Active plans and progress.
Learned Knowledge: Agentic Mode	The agent decides when to save and apply learnings.
Entity Memory: The Four Tools	Entity memory is the agent’s knowledge about the WORLD - the people, projects, companies and systems around the user - as opposed to user memory, which is about the user themselves.
Extraction Limits: Preventing Runaway Loops	Configure max_updates_per_run to cap memory updates per extraction.