1. Home
  2. Workshops
  3. Building Agents with the Claude Agent SDK

Building Agents with the Claude Agent SDK

Over two days your developers build custom agents with the Claude Agent SDK. You start with a minimal agent loop, add tools and permissions, structure work with subagents, and learn how to test and observe agent behaviour. The result is a working agent for a use case from your own context, designed with clear limits and human oversight.

Request without obligation

Deutsche Version

Duration
2 days
Level
Advanced
Format
Remote or on-site, English or German
Group size
Up to 12 participants per trainer

Who it is for

  • Software engineers building agent features or internal agents
  • Teams that want to automate workflows with Claude-based agents
  • Architects evaluating agent frameworks

What your team will be able to do

  • Build an agent with the SDK and explain its core loop
  • Add custom tools and control what the agent may do
  • Use subagents and context management for larger tasks
  • Write tests and evaluations for agent behaviour
  • Apply logging, limits and human approval to a production-minded design
  • Package the agent for use by colleagues

Agenda

  1. Concepts and first agentThe agent loop, models, tools and the structure of an SDK project. A minimal agent runs by the end of the first session.
  2. Custom toolsDefining tools with validated inputs, handling errors and keeping tool output safe for the model.
  3. Permissions and safetyAllow and deny rules, approval hooks, sandboxing and handling untrusted content.
  4. Subagents and contextDelegating subtasks, managing context size and sharing results between agents.
  5. Testing and evaluationBuilding test cases, checking tool use and tracking regressions as prompts and tools change.
  6. Observability and operationsLogs, traces, cost awareness and error handling for long-running agents.
  7. CapstoneParticipants complete and present an agent for a real use case, followed by a review of risks and next steps.

Prerequisites

  • Programming experience in Python or TypeScript
  • Experience using LLM APIs or coding agents
  • API access for the exercises, which we help you plan

Why apprat.io workshops

  • Taught by practising AI Code Engineers
  • Your repositories or a sandbox project
  • AI speed, human responsibility: review, security and governance included
  • ISO 9001 & ISO/IEC 27001 certified processes
  • Follow-up office hours after 30 days

Request without obligation

Tell us about your team and goals. We reply within one business day with a tailored agenda and a quote – free and without obligation.

24/7 call support in English & German:
+49 761 519 751 50

Frequently asked questions

Which programming language do we use?

We use the language you prefer among those supported by the SDK and tailor examples to it.

Do we need API access beforehand?

Yes, access has to be set up before day one. We provide a checklist and can support you.

Is the agent production ready afterwards?

You get a solid foundation and a risk review. Hardening, security review and operations are part of the next steps, which we can support.

Fields marked * are required.

Related workshops

All workshops