AgentPProf brings pprof flame graphs to agent trajectories by segmenting on task boundaries.

arXiv 2609.20301 argues existing observability tools do per-execution debugging but not cross-run profiling, so nobody can answer where failures cluster or which tasks eat the budget at scale. The obstacle is that the responsible entity is a task intent like "diagnose authentication" rather than a code path with a stable identifier, so the authors define a semantic operation stack, recursively split trajectories at task boundaries, and emit pprof-compatible profiles. Segmentation reaches 0.764 B-cubed F1 against human annotations, and the profiles raise problem-localization MAP by up to 56% on three benchmarks.