Every session names its agent, model, and version.
22 of 28 sessions carry the runtime and model
Which agent wrote it, who directed it, who reviewed it. Signed into your own git.
Four facts about the shipped product. The number under each is this repository's own record, right now.
Prices are indicative during early access. Early access teams keep their rate for a year.
A record of the models, versions, and datasets used to generate or test software (3.6.f), review of AI generated code like manually written code (3.6.c(1)), and human review and approval for AI generated changes to security or safety critical functionality (3.6.a(2)). Attestia records the model per session, a named reviewer per agent authored commit, and required review on protected paths. The export is evidence toward the software evidence package the instruction describes.
Which agent, model, and version made each change, which person reviewed it, and that the record has not been altered since the night it was written. It is evidence for controls in the AU, CM, SA, SR, and IA families; the control map says exactly which and what would make it insufficient. We never say compliant.
No. It never adds an approval step. The review that counts is the peer review your team already does; Attestia records it. Only three things are ever blocked, a banned model or tool, a commit with no session identity, and an agent approving its own work, and each has a logged break glass.
Tell us the project and what it has to prove. A person replies within two business days.