Glossary

Agent evaluation

On this page

The short answer

  • Agent evaluation measures a system that takes multiple steps — planning, calling tools, reading results and perhaps trying again — not just one reply. Typical checks include whether it chose the right tool, looped, or finished the task.
  • This is a glossary entry, not a product claim or a benchmark result.
  • All terms · Templates · Guides.

Definition

Agent evaluation measures a system that takes multiple steps — planning, calling tools, reading results and perhaps trying again — not just one reply. Typical checks include whether it chose the right tool, looped, or finished the task.

In practice

Litmiq runs single-turn tests. You can still test the first step (did the model emit the right tool call as JSON?) with the tool-calling template, but you cannot replay a full trajectory here.

Tracing platforms and agent-focused libraries (for example TruLens or Giskard) are built for multi-step runs. Use Litmiq to regress the prompts inside an agent, one turn at a time.

FAQ

Does Litmiq implement “Agent evaluation” as a built-in metric?

Only where the definition itself says so. Many terms are industry vocabulary. Litmiq is a single-turn, BYOK prompt tester with an LLM judge, not a full evaluation platform.

Where is the rest of the glossary?

The hub at /glossary lists every term. This page is the standalone entry for Agent evaluation.

See the term in a suite

Many glossary entries have a matching template you can run in the browser.

  • Free, no signup
  • Runs in your browser
  • Your OpenAI key, sent only to OpenAI

Follow Litmiq on LinkedIn