> ## Documentation Index
> Fetch the complete documentation index at: /docs/llms.txt
> Use this file to discover all available pages before exploring further.

# Experiment, monitor, and export traces

> Run experiments against curated datasets, evaluate production traffic, configure monitors, and export traces to a coding agent.

Improvement doesn't end at launch. Run experiments against curated datasets to prove a change is genuinely better, evaluate production traffic as it flows, configure monitors that surface problems early, and export traces to a coding agent that turns failures into fixes.

## Turn failing traces into a better agent | Ep. 10

[Turn Failing Traces Into a Better Agent | Ep. 10](https://www.youtube.com/embed/4bU7T4s722Y)

Curate a dataset of test cases and run experiments to prove a change is a genuine improvement.

## Run LLM evals on live production traffic with your agents | Ep. 11

[Run LLM Evals on Live Production Traffic with your Agents | Ep. 11](https://www.youtube.com/embed/GLlrNdUIXdQ)

Run evaluations automatically against live production traffic to catch regressions as they occur.

## Catch AI agent failures fast with monitors and alerts | Ep. 12

[Catch AI Agent Failures Fast with Monitors and Alerts | Ep. 12](https://www.youtube.com/embed/XhNosOO8S5I)

Configure monitors and alerts so you detect problems before your users do.

## Build self-improving AI agents with coding agents | Ep. 13

[Build Self-Improving AI Agents with Coding Agents | Ep. 13](https://www.youtube.com/embed/uuZ5DvC7bSA)

Send traces to a coding agent to turn observed failures directly into fixes.

## Up next

> [card] [AI agent mastery](/docs/ax/learn/courses/ai-agent-mastery/overview.md)
>
> A deep dive on building, observing, evaluating, and improving production agents.