In this workshop, you’ll get a beginner’s introduction to OpenAI’s new Agent Builder—why it matters, how it works, and how to build your first agent without writing code.
We walk through the interface, core nodes and tools, and show how you can add real capability with Guardrails, File Search (vector stores), and MCP integrations (Gmail, Drive, Zapier, N8N). You’ll build a working Q&A agent against a knowledge base in about 10 minutes, learn how to choose models and reasoning settings, preview and publish, and get a peak at two of my production-ready workflows: a newsletter agent and a full SEO blog pipeline.
Timestamps
[00:00] Why OpenAI Agent Builder matters (and who it’s for)
[01:20] Where to access the builder (platform.openai.com) and templates tour
[03:00] First look at the builder canvas: nodes, simplicity, and scope
[07:40] MCP explained: Gmail/Drive, third-party tools (Zapier, Shopify, Stripe), and N8N
[09:50] Logic and data nodes: If/Else, While loops, User Approval, Transform, Set State
[11:20] Build-along: creating an “AI Disruptor Archive” Q&A agent
[12:10] Writing system instructions and using the built-in prompt optimizer
[13:30] Choosing models (GPT-5, O1/O3 reasoning) + setting reasoning effort
[14:30] File Search + Vector Stores: upload TXT/JSON, copy ID, connect in your agent
[17:30] Live test: building a basic RAG agent
[21:10] Why this is faster than old stacks (external vector DBs, API keys)
[27:10] Set State pattern: saving and reusing outputs across branches
[28:20] When to use Agent Builder vs. N8N in your stack
