Back to home
Read the Traces docs →
November 18, 2025
See exactly what your AI agent did
Traces are the core of neatlogs. Every agent run is captured as a full span tree — LLM calls, tool invocations, retrievals — with inputs, outputs, timing, and cost.
What you get
Understand exactly what happened during every agent run.
✓ Visualize the full call hierarchy with the span tree
✓ See what ran in parallel, what ran sequentially, and where latency was spent
✓ Inspect prompts, completions, token usage, costs, and timing for every span
Whether you're debugging a failure or investigating performance issues, traces give you the complete execution story.
Quick start
neatlogs.init(
api_key=os.environ["NEATLOGS_API_KEY"],
workflow_name="customer-support",
instrumentations=["langchain", "chromadb"],
)
@neatlogs.span(kind="WORKFLOW")
def handle_request(user_input: str) -> str:
...