Mastering Basics of OpenAI Agent SDK Framework

Learn to build openai agent tools, enable real time weather updates via a weather API, customize model names, and produce structured output by defining an identity class and output type.

Compare local and llm context management in the OpenAI agent SDK, and implement run context wrapper, function tools, and on-demand tools to control what the model sees.

Apply guardrails as a parallel safety net that validates input and output in the OpenAI agent SDK, halting on tripwires when safety criteria fail.

Implement input guardrails on the agent to halt non-visit queries using a tripwire. Define a visit guardrail with a structured output and run context wrapper for syncing and context management.

Enable output guardrail for your agent by implementing an output guardrail and validating output against visit or exploration. The video shows code changes, imports, and tripwire triggers for disallowed outputs.

Explore implementing the model context protocol (MCP) with OpenAI agents by building a custom MCP server and exposing tools, prompts, and resources, using studio or ssk transport protocols.

Configure a local MCP server and MCP client to run custom tools, start an agent, pass server details, and fetch weather data while caching tool lists and tracing executions.

Explore connecting the MCP client to pre-built servers, including the file system and time servers, using Node.js, npm, and Python to read files, run tools, and handle multi-server requests.

Learn to expose an mcp server using sas transport, then connect to a remote mcp server over ssh with a custom server and client to query tools like git.

Adopt streamable http to enable a single bidirectional endpoint for remote MCP servers, replacing the two-endpoint ssh approach, and configure the MCP server and openai agents to connect.

Learn how to integrate open source models with the OpenAI agent SDK, configuring the model at the global, agent, or run level, using llama and grok with chat completion endpoints.

Learn to deploy and connect azure openai with the openai agent sdk by setting up resources, deploying a model, and configuring a client for chat completions.

Explore core patterns in the OpenAI agent SDK framework, including deterministic sequential flow, tool-based routing, hand-off routing, and parallel agents, with input processing, aggregation, and decision routing.

Explore a sequential workflow in the OpenAI agent SDK framework, where a code agent generates Python code, passes it to a validator agent for correctness, edge cases, and error handling.

Explore using agents as tools by registering Spanish, French, and Italian translators, an orchestrator to route tasks, and a synthesizer to inspect translations for accuracy and naturalness.

Master handoff and routing by delegating tasks to Python or Java agents using a handoff description, and verify the selected agent with prints and traces.

Operate parallel German language agents using different models to translate input into German. A judge agent evaluates the translations and selects the best one.