Building Agents with Claude and eve
Build real agents on production harnesses: Claude Code, the Claude Agent SDK and Vercel's eve, with skills, subagents and human approval built in.
2 modules · 6 lessons~20 hoursUpdated 2026-10-30
- Lessons6
- Quizzes6
- Sources37
- Of study~20h
About this course
About this course
You won't write an agent loop from scratch in this course. You'll learn what a harness gives you, then build on two of them: the Claude stack (Claude Code, the Claude Agent SDK, Managed Agents) and eve, Vercel's filesystem-first framework for durable agents. Every concept is grounded in a real document workflow, from reading a pack of documents to routing drafted questions for an expert's sign-off. Exercises run on your Claude subscription, and the eve and SDK exercises on any model through your own OpenRouter key or a free local model.
About the running example
The examples follow a fictional property-law (conveyancing) firm in England and Wales: agents read a contract pack and draft pre-contract enquiries (questions to the other side) for a solicitor to approve. It's a case study, chosen because its documents are long, structured and high-stakes. You don't need a legal background, and the same patterns apply to contracts, insurance claims, invoices, support tickets or patient records. In the lessons, a matter is one client's case and a solicitor is the lawyer who signs work off.
What you'll learn.
- 1Choose between Claude Code, the Claude Agent SDK, Claude Managed Agents and eve for a given agent, and justify the choice with explicit criteria
- 2Build an agent with custom tools, permission rules and hooks in Claude Code, and explain the equivalent Agent SDK code
- 3Author an eve agent from files (instructions, tools, skills, sandbox seeds, subagents) and explain how its durable sessions checkpoint and resume
- 4Package procedures as skills and delegate to isolated subagents, reasoning about what context crosses each boundary
- 5Design approval checkpoints for high-stakes actions and implement them with eve approval policies and SDK permission callbacks
- 6Deploy an eve agent safely: route auth, per-tenant isolation, secrets in the app runtime, and a deployable model
2 modules, 6 lessons.
7h of lessons
Module 1 · 3 lessons · 3h 20m
Harnesses
What a harness gives you, and how the Claude stack and eve package the loop, tools, context and sessions.
Module 2 · 3 lessons · 3h 40m
Composition and control
Skills, subagents and human approval: the pieces that make an agent fit for regulated work, and how to ship it.
A drafting agent with human approval in eve
Build an eve agent that reads a synthetic document pack, drafts follow-up questions with evidence, delegates part of the review to a subagent, and sends nothing until a human expert approves. Runs locally on your own OpenRouter key or a free local model.
More in LLM and Agent Foundations.
- IntermediateAI · Transformers
LLMs for Engineers
How large language models are built, trained and run, so you can reason about their behaviour, cost and failure modes in production.
9 lessons ~18hFree - BeginnerAI agents · Claude Code
Agent Foundations
What an AI agent really is, how it uses tools and context, and how to make one dependable enough for high-stakes work.
6 lessons ~18hFree