AI agents now write most of your code. Your engineering standards didn’t come along. Groundrule puts them in one rulebook, teaches it to Claude Code, Cursor and Copilot, and checks every PR against it, with evidence. Open-source core. Free to start. groundrule.dev

The core is Apache-2.0 and works with no account: CLI, checks, packs, hooks, GitHub Action, MCP server. github.com/Metricall-AI-Lab/… Docs: docs.groundrule.dev

Nothing blocks on day one. Every rule has a stage: observe → teach → advise → enforce. You move it forward when the evidence says it's clean, and a compliance report is built from the actual runs. Rules with proof, not vibes.

How it works, in order: 1. Write a standard once (or run `groundrule scan` and accept what your code already does) 2. `groundrule sync` writes AGENTS.md, CLAUDE.md, Cursor rules and Copilot instructions from it 3. Every PR is checked in CI - standards, the file and line, the fix