Governance for AI-assisted work

Git-backed laws. Deterministic resolver. AI drafts — humans approve.

LawHub is a Next.js workspace on Vercel that lets projects declare local laws, resolve applicable rules on demand, and propose changes through AI — with every commit carrying an audit trailer and landing on a review branch, never on main.

Open workspace Bootstrap a repo

Git is truth

Constitution, local laws, and every approved change live in Git. Runtime KV and Postgres hold only sessions and drafts — never authoritative rules.

Deterministic resolver

A pure function maps (project, scope) to the applicable packet. AI explains — it never adds or drops laws.

Humans approve

Writes go through proposal → review → approve → atomic tree commit on a lawhub/* branch. Default branch is never touched.

AI integration protocol

1Read governance/registry/policy.yml.
2If scope is unclear, call LawHub MCP (https://lawhub.doromo.xyz/api/mcp) and use normalize_intake or resolve_scope.
3Read applicable rules under governance/{laws,methods,protocols,meta}/in the order the task pack returns.
4Begin task (proposal → draft → review → commit).

Governed project layout

governance/
  registry/
    policy.yml
  laws/                # hard constraints & defaults
    common/
      index.md
      reserved-filenames.md
    frontend/
      index.md
      product/
        index.md
        sections/header.md
        overlays/settings/{index.md, danger-zone.md}
  methods/             # reusable solving procedures
    classification.md
    critique.md
    structure-first.md
    …
  protocols/           # staged artifact-gate contracts
    governance/authoring-law.md
    frontend/page-generation.md
  meta/                # governs governance itself
    layer-purity.md
    anti-drift.md
    canonicality.md