Using Microsoft Agent Framework with Foundry managed memory

Microsoft Developer walks through Foundry Agent Service managed memory and how it integrates with Microsoft Agent Framework and LangGraph to build stateful agents that retain context across sessions, including per-user scoping, automatic extraction, and CRUD APIs for memory control.

Overview

Production agents need to remember things across sessions. This episode covers Memory in Foundry Agent Service: a managed, long-term memory layer intended to turn stateless LLM interactions into stateful, continuous agents.

Speakers mentioned in the episode description:

What “Foundry Memory” is for

Foundry managed memory is presented as a way for agents to persist information such as:

The goal is to provide this persistence with zero infrastructure overhead, while still supporting enterprise governance.

Integration points

The episode description says memory is natively integrated with:

Key capabilities called out

The description highlights several developer-facing features:

These are positioned as giving developers control over memory behavior while keeping governance built in.

Video chapters (from description)

Resources (from description)