In this workshop, we will explore agentic AI in the browser; systems that can observe their environment, make decisions, and take action using modern Web APIs. Participants will build autonomous, goal-driven agents that run entirely in the browser, without relying on server-side execution for core behaviour.

Through hands-on exercises, we will implement an agent loop, run agents inside Web Workers, persist state using IndexedDB, and safely expose browser capabilities such as file access and notifications. We will also focus on guardrails: permission models, explainability, user control, and offline-first operation.

By the end of the workshop, attendees will have a working agent architecture and a set of practical patterns they can apply to real-world web applications and internal tooling.