# Agent Trajectory

> The record of each step an AI agent took in one run.

**Field:** Artificial Intelligence  
**Category:** AI Agents  
**Also known as:** Agent Trace, Agent Transcript, Agent Run Log

## Think of it like

It is like the trail of footprints in snow. You can see every step someone took to get where they are.

## Definition

An agent trajectory is the full record of one agent run, step by step. It lists what the agent thought, which tools it called, what came back and what it did next. People read it to find where a run went wrong. Teams also score whole trajectories to test an agent, since a good final answer can hide bad steps. A trajectory is often saved as a trace in an observability tool.

## Example

Emma's agent sent the wrong email. She reads its trajectory and sees that at step 6 a search returned an old address, which the agent then used.

## Why it matters

You cannot fix what you cannot see. The trajectory shows the exact step where a run went off course.

## Common mistake

Checking only the final answer. An agent can reach a right answer by a risky path that fails next time.

## Three principles

- **Every step.** Thoughts, tool calls and results, in order.
- **Find the fault.** Shows where a run went wrong.
- **Test the path.** Score the steps, not just the answer.

## Related terms

- [Agent Loop](https://learn.benibix.com/term/ai/agent-loop)
- [LLM Observability](https://learn.benibix.com/term/development/llm-observability)
- [Distributed Tracing](https://learn.benibix.com/term/infrastructure/distributed-tracing)
- [Long-Running Agent](https://learn.benibix.com/term/ai/long-running-agent)
- [LLM Evaluation](https://learn.benibix.com/term/ai/llm-evaluation)

![A long row of numbered agent steps, read, run, edit, run, running for hours. Under it a log records every step. At the end a stop sign lists goal met, turn limit and budget. Above, a slider runs from suggests only, to asks first, to acts and reports.](https://learn.benibix.com/diagrams/ai7-agent-run.svg)

---

Source: https://learn.benibix.com/term/ai/agent-trajectory
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-19
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
