Podcast: Strands Agents with Clare Liguori

In a recent **InfoQ podcast**, host **Thomas Betts** interviewed **Clare Liguori**, the technical lead on the open source **Strands Agents SDK**. The discussion focused on the operational growth of **Strands Agents** and the architectural shifts required to support systems deployed in live environments.

Technically, **Strands Agents** transitioned from its initial form as a **Python SDK** into a full **agent harness** running in **production**. A core mechanic of this evolution was shifting to a **model-driven architecture**, which redefines how the harness interacts with and leverages the underlying **LLMs**.

For engineers and system builders, moving from simple script-based agent prototypes to scaled infrastructure introduces significant operational complexity. The insights shared from operating **Strands Agents** at scale highlight the necessity of dedicated **agent harness** tooling rather than relying on custom boilerplate code.

In the broader market context, developer tooling for AI integration is advancing from lightweight wrappers to formal execution frameworks. The shift toward a **model-driven architecture** within **Strands Agents** reflects how open source projects are adapting to the increasing capabilities of base **LLMs**.

The practical takeaway for technical teams is to evaluate whether existing custom agent pipelines should be migrated to a production-grade **agent harness**. Moving forward, builders should monitor how changes in the **LLMs** underpinning these systems directly influence the design of the **Strands Agents SDK**.