Agent Memory

Agent memory is how an AI agent retains and recalls information across separate tasks or conversations. Without memory, every interaction starts from scratch.

There are two common types of memory: short-term memory, which is the context included in the current prompt; and long-term memory, which is stored in a database and retrieved when relevant. Most production agents use a combination of both.

The OpsForEnergy agents rely on Supabase as their long-term memory. Project statuses, permit records, and historical actions are all stored there. When an agent wakes up, it queries Supabase for the current state rather than trying to remember everything in its prompt.