This page takes you from nothing to your first LLM trace. It covers tracing the model calls your own application makes, which is the path most people start with. To trace AI coding assistants instead, see Trace your coding assistants.

What you need

Two things, both quick to get:
  • Your Atatus license key, from Settings then License key in the Atatus dashboard.
  • An application that makes LLM calls and runs, or can run, an Atatus APM agent: Node.js with atatus-nodejs 3.6.0 or newer on Node.js 18.19 or newer, or Python with atatus 2.1.0 or newer.

Turn on LLM monitoring

LLM monitoring is an option on the APM agent you already run, so there is no separate service to deploy. Pick your language:
Install the agent and its LLM monitoring package:
Then turn on tracing and LLM monitoring when you start the agent, as the first line of your entry file:
Restart your application. For ES modules, the libraries that are traced, and every option, see Node.js LLM monitoring.

Confirm it is working

Make at least one call through an LLM library your application uses. Then open the LLM page in Atatus. Your application should appear within a minute or two of the first LLM call. It also stays under APM, because it has ordinary request traces as well. If nothing shows up, check these in order:
  1. Look on the LLM page as well as APM. An application appears under LLM only after its first LLM span arrives.
  2. Confirm the agent starts before your LLM library is loaded.
  3. Confirm LLM monitoring is turned on, and for Node.js that tracing is on too.
  4. Raise the agent log level and read the startup logs.
Node.js troubleshooting walks through each of these in detail.

Read your first trace

Open a trace from the LLM page. You are looking at a tree:
  • The request your application handled, at the root, such as an HTTP transaction.
  • LLM spans beneath it, one per model call. Each carries the model name, token counts, and cost.
  • Tool and agent spans where your code or a framework groups calls together.
Span kinds explains every kind you might see, and Cost and tokens explains the spend figures.

Trace your coding assistants

LLM monitoring also traces the AI coding assistants your team uses, such as Claude Code, Codex, and Cursor. That path installs a small tracing package on each developer’s machine instead of changing an application. Each prompt becomes a trace, and these projects appear on the LLM page only. To set it up, follow Installation, or start with the Coding agents overview to see the ten supported tools.

Next steps