Building Production Agents with the Claude Agent SDK

Build a deep-search agent in Python—and learn how to make it more reliable, observable, and ready for real-world use.

In this three-hour live workshop, we’ll build an agent that turns a research question into a plan, searches and reads sources, checks for missing information, and produces a report backed by citations.

Then we’ll address the engineering decisions that matter beyond the first successful run: controlling tool access, managing sessions, handling failures, tracking execution, evaluating results, and keeping costs under control.

I’ll explain the architecture, build the agent step by step, and guide you through practical exercises so you can adapt it yourself.

What you’ll learn

• Build a research workflow with the Claude Agent SDK in Python, using tools, instructions, and structured outputs.
• Improve research quality through evidence tracking, gap checking, and subagent delegation.
• Manage sessions and context so the agent can continue a research task.
• Add permissions, failure handling, execution limits, and useful logs.
• Evaluate the agent’s results and understand the architecture needed to deploy it.

You’ll see how these pieces fit together in one project, with working code you can reuse and extend after the workshop.

Your seat includes

• Three hours of live teaching, guided practice, and Q&A.
• The full workshop recording.
• Starter code and the completed project.
• All presentation slides.
• A 150-page book covering the workshop concepts, implementation, and engineering decisions.

The workshop is designed for people with programming experience who can read basic Python. No previous Claude Agent SDK experience is required. Setup instructions will be shared before the session.

Workshop Details:

• Date & Time: Sun 8 Nov 19:00 - 22:00 EEST
• Location: Zoom Meeting 
• Price: $59 for the first 10 bookings · $79 afterward

Book your seat to build a working deep-search agent and develop the skills to take it further.