Vibe Coding Your First AI Agent with Claude Code

Move beyond simple apps and vibe code a real, working AI agent—entirely from the terminal. In this course, Basia Kubicka teaches you to use Claude Code, the Vercel AI SDK, and a GitHub Codespace to create a sales prospecting agent built on four building blocks: brain, tools, memory, and loop. Connect your agent to Airtable so every action (prospect scores, research results, outreach drafts) lands in a live base you can watch update in real time. Following a test-driven development pattern, you write tests before building each feature so your code is reliable from the start. By the end, you will have a fully functioning agent proof of concept and a clear picture of what it takes to move it to production.

Learning objectives
Design and build a console-based agentic AI system with four core components: a reasoning brain, callable tools, persistent memory, and the loop that ties them together.
Use Claude Code and the Vercel AI SDK inside a GitHub Codespace to scaffold, develop, and test a multistep agent workflow using test-driven development.
Implement the web search API and real-time data output by connecting your agent to Airtable, making its behavior visible and shareable without building a front end.
Distinguish between an AI agent and a workflow by building both, and identify the next steps required to take a proof of concept toward production.