Build AI Agents from First Principles
Hugo Bowne-Anderson
AI Builder/Educator (6 Million Students)
Build the core architecture behind production-ready AI agents
Most agent tutorials stop at a toy demo. This workshop shows you how to build the pieces of an AI agent that can actually do useful work.
You’ll start with a simple LLM call and turn it into a working customer service agent: one that can understand a request, decide what information it needs, use tools, keep track of context, and return an answer you can inspect and improve.
By the end, you’ll have a working agent you can build on, plus a clear mental model for when to use an agent, when not to, and how the pieces fit together under the hood.
Enroll before July 6 to get very early bird pricing at $350. Early bird pricing increases to $425 on July 13, and the full course price of $500 begins on July 20.
What you’ll learn
Build a working customer service agent and learn the foundations behind tool use, harness engineeing, memory, context engineering, and MCP.
Learn how agents really work: LLM calls, tools, state, and reasoning loops
Build an agent in Python before moving to higher-level SDKs and frameworks
Know when to use simple LLM calls, workflows, or full agentic reasoning
Design prompts, tool schemas, memory, and context that guide agent behavior
Use structured outputs and traces to make agent behavior easier to inspect
Add tool feedback and debugging patterns for more predictable systems
Move from a hand-built agent loop to a modern agent SDK
Understand what SDKs handle and what you still need to design
Use MCP to connect your agent to tools and external context
Workshop agenda
From LLM call to agent loop
Start with a simple LLM call, then add tools, state, and control flow so the agent can decide when to act.
Harness Engineering
Design the harness around the model: prompts, tools, context, memory, state, and guardrails, then use traces and feedback to inspect and improve the loop when it fails.
Agent SDKs
Rebuild the core loop with a modern agent SDK and see what the SDK handles versus what you still design.
MCP and reusable architecture
Connect tools through MCP and leave with a starter architecture you can adapt to your own workflows.
Learn directly from Hugo
Hugo Bowne-Anderson
AI & data engineer, consultant, educator of 6+ million students (ex-Yale)
Who this workshop is for
Software engineers, data scientists, and ML practitioners building LLM-powered applications who want a practical foundation for agents.
Technical founders and product-minded builders deciding when an agent is useful versus when a simpler workflow is enough.
Builders curious about agent SDKs, MCP, tool use, and structured outputs who are comfortable lightly editing Python.
Prerequisites
Comfort reading and lightly editing Python
You do not need prior agent experience, but you should be comfortable following code examples and making small changes during the workshop.
What's included
Live sessions
Learn directly from Hugo Bowne-Anderson in a real-time, interactive format.
Lifetime access
Go back to course content and recordings whenever you need to.
Community of peers
Stay accountable and share insights with like-minded professionals.
Certificate of completion
Share your new skills with your employer or on LinkedIn.
Maven Guarantee
Your purchase is backed by the Maven Guarantee.
Frequently asked questions
Maven for Teams
Reimbursement
Get your company to pay
Everything L&D needs: email template, receipts, and certificate of completion.
Get reimbursedTeam discount
Learn with your teammates
Save 20%+ when 2 or more teammates enroll in the same cohort.
Save 20%+ with a teamPrivate cohort
Run a cohort for your org
A dedicated cohort with a custom schedule and curriculum, tailored to your team.
Book a private cohort$350
USD
6–10pm EDT