Select a repository
Pick a repo from the sidebar, choose your scope and two reviewer models, and they'll debate the code and hand you a consensus matrix.
Two AI models debate your code
Pick a repo from the sidebar, choose your scope and two reviewer models, and they'll debate the code and hand you a consensus matrix.
Security
—
The review runs twice with roles swapped (each model audits and defends), then converges on a per-finding consensus. Model identities are withheld from both agents to keep the debate blind.
Pick one or more categories. Every finding still has to carry concrete, checkable evidence — that's what lets the two models adjudicate it.
Two reviewers debate the code. Each one both audits and defends — roles swap across the two runs.