Back to Event Recordings
Workshop Resource

Building AI Agents with MCP, PydanticAI and OpenAI

Build an AI agent from first principles, implement RAG with FAQ data, compare agent frameworks, and expose tools via MCP for reuse across agents and IDEs.

September 1, 2025Intermediate to Advanced
ai-agentsllm-engineeringagent-systemstooling-architecturemcprag

Timestamps

Click any timestamp to jump to that moment in the video

00:00
Welcome, Agenda, and Goals

Introduction to the workshop, agenda overview, and learning objectives

02:14
Setup and Prerequisites (Python, API Keys, uv)

Setting up the development environment: Python, API keys configuration, and uv package manager

05:54
Agents 101 and Tool Demo; Plan for the FAQ Assistant

Understanding agents basics, tool demonstration, and planning the FAQ assistant project

07:50
Parse Google Doc to JSON and Index with MinSearch

Extracting data from Google Docs, converting to JSON, and indexing with MinSearch for search capabilities

14:30
Function Calling with OpenAI Responses API and the Agent Loop

Implementing function calling using OpenAI Responses API and building the agent control loop

22:00
System/Developer Prompts and Multi-Search Reasoning

Crafting effective system and developer prompts, implementing multi-search reasoning patterns

37:00
Testing Workflow with a Simple Runner and Chat UI

Building a testing workflow with a simple runner and interactive chat interface

53:00
Autogenerating Tool Schemas; Adding `add_entry` to Write Back

Automatically generating tool schemas and adding write capabilities with add_entry tool

58:00
Refactor Tools into a Class; Cleaner Design

Refactoring tools into a class-based structure for better organization and maintainability

1:02:00
OpenAI Agents SDK and PydanticAI; Swap to Anthropic

Comparing OpenAI Agents SDK and PydanticAI frameworks, switching to Anthropic Claude

1:17:22
MCP Intro and Building an MCP Server for `search`/`add_entry`

Introduction to Model Context Protocol (MCP) and building an MCP server with search and add_entry tools

1:28:32
MCP Handshake, Listing Tools, and Calling Tools

Understanding MCP protocol: handshake process, listing available tools, and calling tools

1:36:35
Using MCP Tools from the Agent; HTTP Transport Option

Integrating MCP tools into the agent and exploring HTTP transport for MCP communication

1:41:54
Connecting MCP in Cursor and Coding with Live FAQ Context

Connecting MCP server to Cursor IDE and using live FAQ context for coding assistance

1:50:30
Course Overview and Wrap-up

Overview of the AI Engineering Buildcamp course and workshop summary

Core Tools

OpenAI API (Responses API, function calling, Agents SDK)PydanticAIAnthropic APIMinSearchJupyter NotebookuvGitHubFastMCPMCP (Model Context Protocol)Cursor IDE

What You'll Learn

  • Building an agent from first principles with raw OpenAI APIs
  • Implementing function calling and agent loops
  • Indexing and querying FAQ data for RAG
  • Designing tools with schemas and docstrings
  • Refactoring tools into classes
  • Comparing agent frameworks (OpenAI Agents SDK vs PydanticAI)
  • Switching between LLM providers
  • Exposing tools via MCP for reuse across agents and IDEs

Expected Outcome

A working FAQ assistant agent that can search and update a knowledge base, run across multiple agent frameworks, and consume tools via an MCP server usable by other agents and development environments like Cursor

Materials

Prerequisite: Building a Coding Agent (Python/Django)

article

Related Course: AI Engineering Buildcamp

article