Memory has always mattered for personalization and continuity. But as customers move agents from demos into production, another requirement becomes just as important: reliability.
Enterprise teams need agents that not only remember facts, but also apply what they have learned to follow procedures consistently, recover from repeated failure modes, and complete tasks with greater confidence over time. Memory in Foundry Agent Service is built for this shift, with new procedural memory capability, management experiences, and a set of new features such as time-to-live that give developers more visibility and control over what memory stores.
New procedural memory improves agent reliability
In enterprise deployments, a common failure appears quickly: agents often know the right facts and still fail the task because they do not execute the right procedure. They may skip a validation step, misuse a tool, miss a required policy check, or repeat the same flawed pattern on a similar task. Procedural memory is designed to close that gap by helping agents retain and reuse successful execution patterns, so they can complete complex workflows more reliably instead of starting from scratch every time. When used together with agent optimizer in Foundry Agent Service, developers can create self-improving agents by combining design-time optimization of prompts and tools with runtime learning from real task execution.
Procedural memory works in two steps:
- Agent trajectories are ingested and audited to identify successful patterns, inefficient routes, and missing steps. From this, structured procedural memory items are extracted, capturing both “when to use” (task context, preconditions, signals) and “what to do” (ordered actions, required checks, tool usage).
- When the agent encounters similar tasks, relevant procedures are retrieved and injected into the agent’s context, guiding execution with explicit step-level constraints such as required validations, correct tool parameters, and policy en