Agent memory you can query
Agents forget between runs unless memory is designed in. A short tour of how A3T stores what agents learn and lets you query it with AQL.
Agents that run for weeks need to remember what they saw, what they did and what worked. A3T keeps that memory in a store you can query, govern and forget from.
Three kinds of memory
- Episodic: what happened in each run.
- Semantic: facts and entities, linked into a knowledge graph.
- Procedural: the steps that worked, so the agent can reuse them.
Querying it
AQL (Agent Query Language) reads like SQL with graph steps. The docs cover the full syntax.
Governance
Memory is scoped per tenant and per agent, and policy applies to reads as well as writes.