Learn how to break the “lock‑in” myth and supercharge OpenAI’s Agent Builder with real-world tools. This workshop demystifies the Model Context Protocol (MCP) and shows three practical ways to extend your agents: use OpenAI’s native connectors, plug in Zapier for 8,000+ apps, or run your own custom MCP server with n8n.
You’ll see a full, beginner-friendly walkthrough—from setting up an agentic workflow and adding MCP servers, to wiring tools like Airtable and Perplexity, configuring approvals, and handling common hiccups—so you can build reliable, multi-step agent workflows without coding.
Click here to view the virtual whiteboard from this workshop
Here's exactly what you'll learn:
What MCP is and why it unlocks thousands of integrations for OpenAI agents
Three connection paths: OpenAI connectors, Zapier, or your own n8n MCP server
Step-by-step: adding tools (Airtable, Perplexity), setting approvals, and testing
Troubleshooting tips for early-stage bugs and failure retries
How to scale from simple tests to multi-agent, end-to-end workflows
Timestamps Guide
[00:00] Why Agent Builder matters now and the “lock‑in” concern
[00:41] The plan: connect agents to 8,000+ tools (Airtable, Notion, Perplexity)
[02:32] Native tools tour and MCP overview (Model Context Protocol)
[03:44] Official connectors + third-party options (Zapier, Shopify, Stripe, etc.)
[05:03] Strategy: use your own MCP server (n8n) and why
[05:41] Agent Builder UI: creating an agentic workflow and tool slots
[06:30] Zapier path: create MCP server, get API key, connect to Agent Builder
[07:44] Custom path: add your own MCP server (URL, label, access key)
[08:00] n8n setup: MCP Server Trigger and adding tools (Airtable example)
[09:41] Search/update Airtable records; grabbing the production MCP URL
[10:27] Connecting the MCP server in Agent Builder + approval settings
[11:52] Adding Perplexity via n8n and letting the model define parameters
[13:33] Agent instructions: research assistant demo and preview run
[15:58] Tool approval payloads, retries, and common failures
[18:12] Result: 3 recent news stories via Perplexity (links included)
[19:04] Next steps: scale to multi-agent workflows; start simple and expand
— Alex
Founder of AI WriterOps
Founder of AI Disruptor
