0G Lab, in collaboration with teams from the National University of Singapore, Peking University, and Beijing University of Posts and Telecommunications, has built a consensus protocol debugging agent framework called Agora

Odaily Planet Daily News: 0G Labs, in collaboration with research teams from the National University of Singapore, Peking University, and Beijing University of Posts and Telecommunications, proposed the first automated testing framework, Agora, in their accepted paper at ICML 2026. This framework integrates domain knowledge with LLM-based multi-agent collaboration. Through the interaction of three specialized agents—a Coordinator, a Strategist, and a Code Officer—it drives an automated testing architecture.

In tests on consensus protocol libraries such as Raft, EPaxos, HotStuff, and BullShark, Agora successfully discovered 15 previously unknown deep-level logical vulnerabilities at the protocol level. In comparison, native large models like GPT-5.2 and Claude 4.5 failed to identify any vulnerabilities. Evaluation results show that 73.9% of the vulnerability reports generated by Agora contain genuine logical flaws. On average, discovering one logical vulnerability consumes approximately 5.32 million tokens, equivalent to about $40. The framework's code has already been released in the open-source community.