Kitaru Workshop by ZenML

​Want to take an agent that works on your laptop and run it as a service your platform team can actually operate?

​Most teams are either running agents locally in a notebook, or locked into some closed platform that ties them to one model, one harness, one cloud. Today, it's increasingly obvious that every company needs its own internal agent runtime - one that runs the harness they already picked (PydanticAI, Claude Agent SDK, OpenAI Agents SDK, raw Python), recovers from crashes, can wait days for a human approval without burning compute, and executes risky tools in isolation.

​In this workshop, we'll uplift a Python agent into a durable service on real infrastructure - state that survives crashes, replay from any checkpoint, human/agent approval gates, isolated tool execution - in a way that a platform team can operate without babysitting every run. We'll use Kitaru, the open-source runtime ZenML has been building.

​Bring a laptop with Python 3.10+