Build AI agents with the Claude Agent SDK, connect them to live business systems and deploy them safely as a developer or technical lead
2 day course 10am - 5pm Live Online or Classroom-based
What you will learn
- Understanding how AI agents differ from automated workflows, enabling you to choose the right approach for each task.
- How to build a working Claude agent in Python from first principles, giving you full control over it.
- Designing reliable tools for Claude, helping your agent choose the right action first time and avoid costly loops.
- How to build agents faster with the Claude Agent SDK, using built-in tools for files and web search.
- Connecting agents to business systems through MCP, empowering them to automate real work such as processing incoming orders.
- How to keep long-running agents focused using Skills and subagents, improving productivity on complex multi-step tasks.
- Controlling what your agent can do with hooks and approvals, ensuring actions like payments need human sign-off.
- How to test, measure and deploy agents with Claude Managed Agents, delivering dependable automation your organisation can trust.
Related Courses
This Building Claude Agents course includes
-
Authorised Certificate
-
Lifetime support after your course
-
Course materials
-
Live Instructor-led training
-
Refreshments all day
-
Latest Mac or PC ready to go
-
Authorised Trainers
-
Hands-on exercises
Book this course
We offer Building Claude Agents as a private course.
You can select a date which suits you or your team.
The course content can be fully customised to your exact learning requirements.
Request datesVideo success stories
Course Outline for Building Claude Agents
Understanding What an AI Agent Is
- Understanding the difference between AI agents and automated workflows
- Exploring how agentic AI systems use tools in a loop to decide their next step
- Recognising when a simple prompt or workflow is a better choice than an agent
- Comparing ways to build on Claude: the Claude API, the Claude Agent SDK and Claude Managed Agents
- Planning an agent around a clear goal, boundaries and a definition of done
Building Your First Agent Step by Step
- Setting up a Python environment and a Claude API key
- Defining tools with names, descriptions and JSON input schemas
- Handling tool use and tool result messages in the Claude Messages API
- Writing a simple agent loop in Python and understanding stop reasons
- Setting iteration limits and timeouts so an agent fails safely
Giving Your Agent the Right Tools
- Writing tool descriptions that help Claude choose the right tool every time
- Choosing the right size of tool for each job
- Returning clear results and error messages an agent can act on
- Testing and improving a toolset by comparing agent runs
- Avoiding common tool design mistakes that cause loops and wrong actions
Working with the Claude Agent SDK
- Installing the Claude Agent SDK and understanding how it relates to Claude Code
- Running one-off queries and multi-turn agent sessions
- Using built-in tools for files, shell commands and web search
- Adding custom tools as in-process MCP servers
- Setting system prompts, CLAUDE.md project instructions and structured outputs
- Resuming sessions and keeping a record of agent activity
Connecting Agents to Your Data and Systems
- Understanding the Model Context Protocol (MCP) and how agents use it
- Connecting an agent to existing MCP servers for databases, documents and business applications
- Creating a simple MCP server in Python
- Managing API keys, credentials and secrets securely
- Deciding what an agent should be able to read and change
Keeping Long Tasks on Track
- Managing the context window during long-running agent tasks
- Deciding what belongs in prompts, tool results and files
- Using compaction, summaries and session memory
- Packaging repeatable procedures as Agent Skills with SKILL.md
- Delegating focused work to subagents and running tasks in parallel
- Knowing when multi-agent designs add cost without adding value
Controlling What Your Agent Can Do
- Setting permission modes and allowed tools
- Using hooks to validate, log or block tool calls
- Adding human approval before irreversible actions
- Recognising and reducing prompt injection risks from web pages, emails and documents
- Designing agents that are auditable and easy to stop
Testing and Improving Agent Performance
- Building a test set of realistic tasks with known outcomes
- Grading agent output with code checks and model-graded rubrics
- Tracing agent runs to find where things went wrong
- Measuring cost, speed and success rate per task
- Re-testing whenever models, prompts or tools change
Putting Your Agent into Production
- Understanding Claude Managed Agents: agents, environments, sessions and events
- Running agents in Anthropic-managed cloud sandboxes or self-hosted sandboxes
- Streaming results and steering a running agent session
- Checking data retention, security and compliance requirements before launch
- Choosing between the Claude Agent SDK and Claude Managed Agents for your project
- Monitoring and maintaining agents after deployment
FAQs about the Building Claude Agents course
Is this Building Claude Agents course suitable for me?
This course is designed for developers and technical professionals who are comfortable writing Python and calling web APIs. Experience of Claude Code is helpful but not essential. If you have any questions about whether the course could meet your needs, please speak to our customer service team.
How is the content taught throughout the Building Claude Agents course?
Our Building Claude Agents course is extremely interactive. You will be using Claude throughout the day, working through examples and live projects under the guidance of our trainer, who will be inviting questions and supporting you through each step.
How many people are likely to be in the class when I attend the Building Claude Agents course?
For private classes in Building Claude Agents, we recommend a maximum of 8 to 10 people from the same company in the class. Please ask our customer service team to help you find the right balance of people per class and the number of days training you need.
Should I attend the Building Claude Agents course in person or live online?
For in-person courses at our London training centre, you will receive a warm personal welcome, your choice of the latest Mac or PC and morning and afternoon refreshments. If you prefer to attend remotely, rest assured that you will be taught live by the same experts that deliver our in-person courses, covering the same training content, using the same exercises and materials. A second monitor is recommended but not essential.
What support is available after I attend the Building Claude Agents course?
We offer ongoing support to our Building Claude Agents course attendees through email and by arrangement, phone calls. You are welcome to ask any questions on topics covered during training. Our trainer will always do their best to explain principles and help you with any issues you encounter.
Will I receive a certificate upon completion of my Building Claude Agents course?
Yes, you'll receive a certificate of completion from Transmedia, demonstrating your knowledge and attendance on our Building Claude Agents course. Our certificates note our Authorised Training Centre status.
Building Claude Agents — Team and Bespoke Training
Booking a private training session allows you to focus on your exact needs on dates that work best for you or your team.
- Sessions can be run online or in person, on your site or at our London training centre.
- We manage training for many of the UK’s largest companies, taking care of all the admin, working with you to ensure the perfect course.
- Our expert trainers will first understand your requirements and then tailor each session to your exact specification.
- We offer a free consultation service to work out every detail of your course.
Private One-to-One Training
One-to-one tuition tailored to your exact requirements. Your choice of duration, dates and content.
Book NowTeam Training for Companies
Group bookings provide you with the most cost-effective way to train your team.
Book Now
Our Clients
Latest Blog Articles
Claude Connectors: What Every Creative Professional Should Know
Google Guide to Generative AI: Tips for UX Designers and Content Teams
Generative Engine Optimisation: Practical Tips for Better OutputWhy Transmedia?
Unlimited Training Support
We offer unlimited training support after your course.
Bespoke Training Solutions
Customised training and software solutions based on your exact requirements.
Worldwide Creative Skills Training
Join our authorised training courses live online from anywhere in the world.
Authorised Training Professionals
Our experts are accredited by Adobe, Apple and Claris and bring years of experience and real-world expertise.