Governance that lives in the codebase,
not in pdfs no one reads.
GovKit generates three artifacts your AI coding agents (Cursor, Claude Code, Codex) use during execution: a govkit.md governance file, an evidence manifest, and a Python-based deterministic compliance check. Regulatory controls become project context. Compliance happens at the moment code is written, not after the audit arrives.
The problem
Governance lives in the wrong place
Inside most companies, compliance is a cost center: a legal problem, an engineering ticket that never clears the backlog. So governance gets bolted on at the end, or skipped entirely. Users feel that.
The resolution
Teach the AI what good governance looks like
Encode your values as policy. Run deterministic tests automatically. When governance speaks a language AI understands, it stops being a human burden and starts being a system property: baked in from line one, present in every decision, invisible to the builder, felt by the user.