DeepAudit

Why we included this project

DeepAudit runs a code audit as an autonomous multi-agent workflow rather than a fixed set of pattern-matching rules. It coordinates four agents that trace cross-file calls and business logic the way a human security expert would, then adds a verification step: the verifier agent builds and runs proof-of-concept exploits in isolated containers to confirm whether a finding is a real bug or a false positive. Because it can use local models through Ollama, the whole audit can happen without source code leaving your network. That makes it a practical option for devsecops teams doing pre-release deep reviews, and for smaller teams that want verified findings instead of a pile of unconfirmed warnings.