Traces

The Neatlogs trace screen is where the magic really happens. We’ve poured a lot of love into making it the most refined, easy-to-understand way to look at an agent trace. You don’t need to be a hardcore developer to make sense of it.

Think of a trace as a story: the complete timeline of every step your AI agent (or a whole team of them) took to get a job done.

Here are the components of the trace screen:

Top Bar

• Left: Shows the trace path (Project Name / Trace ID). Use navigation buttons to move between traces.
• Right: Buttons to expand or collapse all agent steps within the trace, allowing for quick overviews or detailed views.

Agents List (Left Panel)

• Lists all agents involved in the trace.
• Shows the number of iterations (steps) each agent took.
• Click an agent's name to scroll directly to its part of the trace.

Trace Details (Left Panel, below Agents List):

• Trace Response Button: Click to jump directly to the final output of the entire agentic process.
• Evaluation Area: Provide or view the evaluation for this trace.

The Trace

This is where all the action (read: debugging) happens. Here you get to see each agent’s detailed, step-by-step execution. Each agent's actions are organized chronologically. These steps or actions include:

• Thought: The agent's reasoning before acting.
• Error: Highlights the errors the model has run into while performing a task.
• Using Tool: Shows the agent calling a tool, including tool name, input, and output. Tasks delegated to other agents also appear here.
• Agent Response: The individual response generated by a specific agent.
• Trace Response: The final overall output of the entire program/automation.