Integrations

Architect 1.0 targets Claude Code. Cursor, Windsurf and GitHub Copilot output writers were removed — if you need them, pin 0.7.13.

GitHub Actions

Six lines. Fails the build on any critical violation, because check exits 1.

name: architecture
on: [push, pull_request]
jobs:
  check:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-node@v4
        with: { node-version: 20 }
      - run: npx @levironexe/architect check .

Ratcheting an existing project

A codebase with existing violations should not go red on day one. Commit a baseline, then fail only on regressions:

      - run: npx @levironexe/architect verify . --strict

Generate the baseline once and commit it:

npx @levironexe/architect check . --baseline
git add .architect/baseline.json

Claude Code

The Stop hook

This is the piece worth setting up. It runs check after every agent turn, so the agent sees what it broke before you do.

In .claude/settings.json:

{
  "hooks": {
    "Stop": [
      {
        "matcher": "*",
        "hooks": [
          {
            "type": "command",
            "command": "npx @levironexe/architect check . --json"
          }
        ]
      }
    ]
  }
}

The JSON lands back in the session, so the agent can read the file, line and fix fields and correct itself without you having to notice first.

The skills

npx @levironexe/architect init .

Installs three skills into .claude/skills/:

The skills read the blueprint's description, bad_example and good_example fields — including for auth_mechanism_mismatch, which check never reports. The agent gets more guidance than the linter enforces, by design.

Any other agent

There is no dedicated writer, but check --json is a stable contract. Any tool that can run a command and read JSON can consume it:

npx @levironexe/architect check . --json | jq '.violations[] | "\(.file):\(.line) \(.fix)"'