​In this hands-on workshop, we'll instrument a Next.js agentic application with behavioral tracking across three layers - client, server, and the agent's own reasoning - and then feed real-time user context from Snowplow Signals back into the agent's system prompt using the Vercel AI SDK.

​

​We'll cover the following steps:

• ​Set up a Snowplow pipeline and validate events against schemas in real time
• ​Add client-side, server-side, and agent self-tracking to a travel chatbot - capturing what the user did, what the agent did, and what the agent decided
• ​Configure Snowplow Signals to compute live session attributes from a behavioral event stream
• ​Inject those attributes into an AI agent's system prompt so responses adapt to what the user is actually doing on the page

​By the end of this workshop, you'll be able to track what your AI agent does at every layer and feed live behavioral context back into its responses using Snowplow and the Vercel AI SDK.

‍