saphan-ai/proven is a Go code reviewer designed to review code changes and report only issues it can prove. It only files a report when there is a test that reproduces the bug or data that confirms the risk. The repository does not present a latest release version. It is released under the Apache-2.0 license, and according to the repository description it was created by Saphan Studio and several AI agents for Open Agent Hackathon 2026.
The core of the tool is that it does not assert a problem on suspicion alone, but requires evidence that can be checked repeatedly. The README introduces command-line inspection and dependency evidence-collection cases alongside code change review. In its security review, it divided the code into several boundary areas, examined them independently and compared the results, and records that the weaknesses it found were confirmed by tests and then fixed. It also includes examples that run without network access and credentials. This material shows how the review process and its evidence are handled.
However, what the README describes is a hackathon development log, step-by-step run records and case studies. The repository description gives no information about actual deployment or scope of use, and no latest release is recorded either. So the strength that can be confirmed at present is the evidence-requiring review method and the material showing that process, while the basis for judging product maturity beyond that is limited.