AgentDebugX: An Open-Source Toolkit for Failure Observability, Attribution, and Recovery in LLM Agents

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Kunlun Zhu, Xuyan Ye, Zhiguang Han, Yuchen Zhao, Bingxuan Li, Weijia Zhang, Muxin Tian, Xiangru Tang, Pan Lu, James Zou, Jiaxuan You, Heng Ji
• Venue: arXiv: cs.AI
• Published: 2026-07-21
• Collected: 2026-07-22
• URL: http://arxiv.org/abs/2607.18754v1
• PDF: https://arxiv.org/pdf/2607.18754v1
• DOI: UNKNOWN

Plain-Language Brief

AgentDebugX is an open-source toolkit designed to aid in the observation, identification of root causes, recovery, and rerunning of issues in LLM (Large Language Model) agents. The toolkit, DeepDebug, uses multi-turn diagnosis through global trajectory understanding, structure-guided investigation, and cross-examination. It outperforms other methods on the Who and When benchmark and GAIA, leading to improved overall accuracy.

What the Document Says

• AgentDebugX is an open-source debugging framework
• It organizes debugging as a closed loop of Detect, Attribute, Recover, and Rerun
• DeepDebug performs multi-turn root-cause diagnosis
• Global trajectory understanding, structure-guided investigation, and cross-examination are used for diagnosis
• On the Who and When benchmark, DeepDebug achieves the best strict attribution accuracy compared to evaluated methods on tested open-weight backbones
• DeepDebug reaches 28.8 percent exact agent-and-step accuracy on qwen3.5-9b versus 21.7 percent for the strongest single-pass baseline
• On GAIA, DeepDebug repairs 13 of 73 failed tasks in a single rerun, compared with 4 to 6 for three decoupled self-correction baselines
• Overall accuracy improves from 55.8 percent to 63.6 percent using AgentDebugX on GAIA
• The toolkit is exposed through a Python library, CLI, web console, and installable agentic skill
• An opt-in Error Hub is provided for sharing scrubbed failure-diagnosis-repair bundles and reusing them as debugging memory

Key Concepts Mentioned

• AgentDebugX
• DeepDebug
• LLM agents
• root cause diagnosis
• Who and When benchmark
• GAIA

Claims Made by the Document

• AgentDebugX improves failure observation, attribution, recovery, and rerunning in LLM agents
• DeepDebug outperforms other methods on the Who and When benchmark and GAIA
• DeepDebug achieves the best strict attribution accuracy among evaluated methods on tested open-weight backbones

Evidence Used

• AgentDebugX organizes debugging as a closed loop of Detect, Attribute, Recover, and Rerun
• DeepDebug uses multi-turn root-cause diagnosis through global trajectory understanding, structure-guided investigation, and cross-examination
• On the Who and When benchmark, DeepDebug achieves 28.8 percent exact agent-and-step accuracy on qwen3.5-9b versus 21.7 percent for the strongest single-pass baseline
• On GAIA, DeepDebug repairs 13 of 73 failed tasks in a single rerun, improving overall accuracy from 55.8 percent to 63.6 percent

Methods / Approach

UNKNOWN

Limitations Stated by the Authors

UNKNOWN

Limitations Noted by the Agent

Agent interpretation — not from the source.

UNKNOWN

Open Questions

UNKNOWN

Notes for Future Q&A

May include interpretation. Labeled where it does.

UNKNOWN

Raw Abstract / Excerpt

LLM agent failures are difficult to debug because the step where an error surfaces is often not the one that caused it. Existing observability tools replay execution traces but provide little support for identifying the root cause or translating diagnosis into recovery. We present AgentDebugX, an open-source debugging framework that organizes debugging as a closed loop of Detect, Attribute, Recover, and Rerun. At its core, DeepDebug performs multi-turn root-cause diagnosis through global trajectory understanding, structure-guided investigation, and cross-examination. On the Who and When benchmark, DeepDebug achieves the best strict attribution accuracy among the evaluated methods on both tested open-weight backbones, reaching 28.8 percent exact agent-and-step accuracy on qwen3.5-9b versus 21.7 percent for the strongest single-pass baseline. On GAIA, DeepDebug repairs 13 of 73 failed tasks in a single rerun, compared with 4 to 6 for three decoupled self-correction baselines, improving overall accuracy from 55.8 percent to 63.6 percent. AgentDebugX exposes this workflow through a Python library, CLI, web console, and installable agentic skill, and provides an opt-in Error Hub for sharing scrubbed failure-diagnosis-repair bundles and reusing them as debugging memory.