Agents in Prod | Amsterdam

To secure your spot, sign up on this page here.

Join the Agents in Prod community for an evening of hands-on building, live demos, networking, and discussions in Amsterdam. This event is brought to you by GlassFlow and Motherduck.
​The technical demos and discussions will revolve around how to build production-grade AI Agents and what to consider when moving your agents from the demo stage to production.

​What to expect:

• ​Technical discussions on building production-grade AI Agents
• ​Hands-on demos and building with GlassFlow and Motherduck engineers.
• ​Food, peer connections, and optional attendee demos.

​Agenda

• ​Agenda

• 18:00 - 18:30: Welcome, arrivals, and registration

• 18:30 - 18.50:
  Building an agent-native datastore with DuckDB
  How GlassFlow built Tares, an open source platform for always-on agents, on a single DuckDB file, and what agents in production actually need from a datastore. The talk includes a demo of cross-source correlation, agent-derived SQL views, and a trigger that wakes an agent with the full context attached.
  Speaker: Ashish Bagri Co-founder, CTO @ glassflow.ai

• 18:50 - 19:15: Q&A, Networking Break, pizza and drinks

• 19:15 - 19:35:
  How to Enforce Agentic TDD with MCP
  Prompting a coding agent to follow TDD (test driven development) doesn’t guarantee that the agents will adhere to the workflow.
  This talk uses an open-source agentic TDD project to show how MCP tools make each state in TDD explicit and deterministic for AI. We'll show a practical pattern for placing workflow invariants in tools while keeping human intent and approval in control for shipping software.
  Speaker: Nnenna Ndukwe, AI Developer Relations Engineering Lead, Qodo

• 19.35 - 19:55: Q&A, Networking Break, pizza and drinks

• 19:55 - 20.15:
  Principles for building an Agent Interface
  Αgents have become a common way to interact with infrastructure products. How do you know whether an interface actually works well for an agent?
  This talk is about the eval system we built to answer that question for our CLI. We test whether an agent can find the right tools, use them correctly, compose them across a task, and do so across different kinds of user requests and personas. From those evals, we started to see a set of principles for building interfaces that agents can use reliably, and this is what I’ll share.
  Speaker: Adithya Krishnan, Software Engineer AI @ MotherDuck

• 20.15 - 21.00: Q&A, Networking Break, pizza and drinks

​Bring your laptop and an idea you want to build.

​Come for the technical deep dives. Stay for the conversations, connections, and drinks.

​Free to attend. Limited spots available. Register now