Agent Cookbooks

Tracing and Evaluating Agents

Agents Cookbook

Build a customer support agent to trace activity, assess performance, and experiment with prompts and models.

Evaluate an Agent

Trace and evaluate a "talk-to-your-data" agent. Includes evaluations for function calling accuracy, SQL query generation, code generation, and agent execution path.

OpenAI Agents SDK Cookbook

Create an agent with the OpenAI Agents SDK, trace its activity, benchmark with datasets, run experiments, and evaluate traces in production.

Evaluating Agents with Ragas

Create a customer support agent using the OpenAI Agents SDK, trace its interactions, and evaluate performance using Ragas.

Tracing and Evaluating Amazon Bedrock Agents

Build an Amazon Bedrock agent, instrument and trace it with Phoenix, and add evaluations to your agent traces.

Tracing and Evaluating a LangChain OpenAI Agent

Build your own LangChain OpenAI agent using the function-calling API and inspect the agent's internals—all in a minimal setup with conversation and tool use.

Tracing and Evaluating a LlamaIndex OpenAI Agent

Use the function-calling API to create a LlamaIndex OpenAI agent capable of conversation and tool use, and explore its behavior with Phoenix.