# Alyx gets better at doing things

> Source: <https://arize.com/blog/alyx-gets-better-at-doing-thingss/>
> Published: 2026-11-20 17:34:00+00:00

**Key takeaway:** Alyx can now create, update, pause and resume Managed Agents and create or edit Decision Judges, including their evaluation questions, classification choices and variable mappings, from within Arize AX.

We’ve expanded what [Alyx](https://arize.com/docs/ax/alyx), the AI engineering agent inside Arize AX, can do. Alyx now works across more of the AI engineering workflow, helping you investigate production issues, configure agents and evaluations, and visualize performance without navigating between tools.

Here’s what’s new.

## Investigate issues found by Signal

[Signal](https://arize.com/docs/ax/observe/signal) continuously analyzes production traces to find recurring issues in your AI applications. Now you can ask Alyx about those issues, inspect the underlying traces, and dig deeper into what went wrong.

Ask what’s happening in a project, explore specific failures, and get recommendations for what to investigate next.

## Configure Managed Agents

You can now create, update, pause, and resume [Managed Agents](https://arize.com/docs/ax/agents/build-your-own-agent) directly through Alyx.

Describe what you want an agent to investigate, define when it should run, and Alyx helps configure it. Managed Agents then run independently, investigating issues and taking actions based on their configuration.

## Create Decision Judges

Alyx can now create and edit [Decision Judges](https://arize.com/docs/ax/evaluate/decision-judge#decision-judge), including their evaluation questions, classification choices, and variable mappings.

Describe what you want to evaluate, and Alyx helps set up the judge. Once configured, you can use it in Arize AX evaluation workflows to assess agent behavior.

## Build and explore dashboards

Create [dashboards](https://arize.com/docs/ax/machine-learning/machine-learning/how-to-ml/dashboards#dashboards), add widgets, change metrics and filters, and explore your data using natural language.

You can also ask Alyx to explain what a dashboard is showing, making it easier to understand trends in latency, cost, and evaluation results without manually configuring every chart.

## Try it in Arize AX

Open Alyx and try asking:

- “What issues has Signal found this week? Help me investigate the most important one.”
- “Create a Managed Agent that investigates spikes in tool-call latency.”
- “Create a Decision Judge that checks whether my agent escalated correctly.”
- “Build a dashboard showing p95 latency by model.”

These updates make Alyx a more capable way to work across Arize AX, connecting production investigations, agent configuration, evaluations, and performance analysis in one place.
