Keep what your agents learn about the project, inside the repo.
Grounded in available product and source data
Naetive captures what your agents learn about a project as grounded claims that point at the code they came from. Claude Code, ChatGPT and Codex all read the same picture, and each one leaves it stronger.
When the code moves past a claim, the tool flags it for review instead of letting it drift. You can export the whole thing as a plain .naetive folder in your repository, so a fresh clone starts already informed.