Clare Liguori is a Senior Principal Software Engineer at Amazon Web Services working on developer tooling and agentic AI. Her work includes Kiro, the Strands Agents SDK and core maintenance of the Model Context Protocol specification.[1]
Start with Introducing Strands Agents, an Open Source AI Agents SDK →
Education
Links
Personal biography
X profile 
GitHub · clareliguori Senior Principal Software Engineer · Amazon Web Services
Career
Liguori joined Amazon in 2014 after software engineering work on WiFi products at Motorola Solutions. Her AWS work has included Proton, ECS and the Code Suite, alongside open-source projects such as AWS CDK, Spinnaker and Backstage. Her current focus combines developer tooling with agent execution and protocol infrastructure.[1]
Contributions
Her Strands launch article describes a model-driven agent loop: define tools and instructions, then let the model decide which operation to invoke and use its result in the next step. The SDK packages model access, tool execution and iteration so an application can begin with a small amount of orchestration code. Supporting production deployment still requires observing the loop and constraining the tools available to it.[3]
Ideas
In her frontier-development talk, Liguori explains how a team prepares repositories and feedback loops for agents doing larger units of work. Clear task boundaries, tests and ways to inspect results let a team assess a change before relying on it. The engineering question is which parts of the development process can check and correct their own output as human supervision becomes less continuous.[4]
Tools, writing & talks
Writing, research & talks
Start here
Introducing Strands Agents, an Open Source AI Agents SDKaws.amazon.comA first-party introduction to a model-directed agent loop and its tool interface.
Follow the smallest agent-and-tool example before the deployment material. Python and familiarity with model APIs help; inspect which tools the example exposes and how its execution is observed.
Author[3]Amazon watched 50 ordinary teams for most of a year. Same assistant, wildly split outcomes: under 3x for half, a 4.5x median for the rest. Liguori defines frontier development by behavior and lists five habits, including doing real brownfield prep work before agents succeed there.
Catalog workOriginal sourceA talk about preparing teams, repositories and feedback for agent-driven development.
Presenter[4]