foundry-memory

What problem does it solve?

Foundry agent teams need a reliable way to remember user-specific preferences and cross-session conversation context without bolting on external memory middleware.

Core Features & Use Cases

• User profiles: persist stable facts and preferences about a user (for example, UI choices, units, role context).
• Chat summaries: store distilled summaries of prior threads so a later conversation can resume without replaying the full transcript.
• Agent-native tool wiring: attach the Foundry Memory tool so the agent can automatically retrieve relevant memories and write new ones after inactivity using scope isolation.

Example use case: build a hosted Foundry agent that personalizes responses for each user by recalling their stored preferences and maintaining a rolling summary of the discussion over multiple sessions.

Quick Start

Configure a Foundry Memory Store with your chat and embedding model deployments, attach the memory tool to your agent with scope set to {{$userId}}, and then have the agent use stored memories while updating them after inactivity.