The 4 Layers of Agent Memory
1. Chat (short-term, same session)
When you tell your agent something in the chat, they'll use that information for the conversation you're having right now. This is great for quick context, but once the session ends, it's not automatically saved.
Use chat for: Same-day instructions, quick context, one-off requests.
2. Tasks (scheduled reminders with context)
When you create a task, you're not just scheduling an action, you're giving your agent a reminder with the full context they need to carry it out. At the scheduled time, they'll receive that context and act on it.
Use tasks for: Anything that needs to happen later, on a schedule, or repeatedly.
3. Brain (permanent knowledge)
The Brain is where your agent's long-term knowledge lives. Anything you add here — memories, documents, web pages, or images, stays available to your agents for every future interaction.
Use the Brain for: Business information, brand voice, pricing, client preferences, recurring instructions, and anything your agents should always know.
4. Guidelines (behavioral rules)
Guidelines are the current rules that shape how each agent behaves within their specific function. They control tone, process, and priorities. You can adjust them by chatting with your agent, or fine-tune them manually in their settings.
Use guidelines for: Defining how your agent should work, not just what they should know.
What can't agents change on their own?
Most settings, like connected accounts, integrations, and plan configurations, need to be changed manually by you. Your agents can update guidelines through chat, but they can't always modify their own settings.