# Trace

> A record of model calls, tool calls, inputs, outputs, timings, and decisions during an AI run.

- Category: Evaluation & Safety
- Stage: Rising
- Also called: Run trace, Execution trace
- Canonical: https://www.joelcsaji.com/ai-dictionary/trace

## Plain English

A trace is the flight recorder for an AI workflow.

## Example

A developer inspects which tool call caused a bad support answer.

## Why it matters

Traces make agent behavior debuggable, auditable, and measurable.

## Related

- [Evals](https://www.joelcsaji.com/ai-dictionary/evals)
- [Tool Result](https://www.joelcsaji.com/ai-dictionary/tool-result)
- [Human-on-the-Loop](https://www.joelcsaji.com/ai-dictionary/human-on-the-loop)

## Mentioned in

- [The AI-Powered Developer: Building Modern Web Apps in 2026](https://www.joelcsaji.com/blog/ai-powered-developer-2026)
- [The Evolution of Agentic AI: From Origins to Future Horizons](https://www.joelcsaji.com/blog/evolution-of-agentic-ai)

## Sources

- [OpenAI: Function Calling](https://developers.openai.com/api/docs/guides/function-calling)
- [Anthropic: Effective Harnesses](https://www.anthropic.com/engineering/effective-harnesses-for-long-running-agents)