Enferno is a Flask framework your coding agent already understands: one way to do each thing, no build step, and checks it runs on its own work.
Open Claude Code, Codex or Cursor in an empty folder and paste this. Replace the last line with your app.
Set up github.com/level09/enferno in this folder and follow its AGENTS.md.
Then build: a customer portal where clients see their invoices.
Run checks.py when you're done and show me the result.Already have an agent setup? Add the skill: npx skills add level09/enferno
git clone [email protected]:level09/enferno.git && cd enferno
./setup.sh # deps + .env with generated secrets (--full for Redis/Celery)
uv run flask create-db # tables, migrations stamped
uv run flask install # admin user
uv run flask run # http://localhost:5000Agents go wrong when a codebase has five ways to do one thing. Enferno has one, writes it down, and lets the agent check its own work.
A 90-line AGENTS.md covers the stack, routes and hard rules. The enferno-dev skill holds the full code patterns.
checks.py runs real code paths against your database. The agent runs it, reads it, fixes what broke.
Vue 3, Vuetify and Tabler icons run straight in the browser. No webpack for the agent to fight.
Everything is a CLI call an agent can run and read.
SQLite by default, so the first run needs no infrastructure. Add the rest when you decide.
./setup.sh --full
Flask-Security-Too with TOTP 2FA, passkeys, recovery codes, Google and GitHub login, roles, and CSRF on every form and API call.
docker compose up --build brings up the full stack.
The same agent that writes the feature verifies it and gets it to production.
Describe the feature in plain words. The agent reads AGENTS.md first.
"add invoices for clients"
Model, migration, route and page, following the skill's patterns.
flask db migrate
It runs the smoke checks and the linter, then fixes what broke.
python checks.py
Docker Compose, or a fresh Ubuntu VPS with Ignite.
docker compose up
Factory pattern, blueprints and SQLAlchemy 2.x. Code you can read in a pull request.
from flask import Blueprint, render_template
from flask_security import auth_required
portal = Blueprint("portal", __name__)
@portal.before_request
@auth_required("session")
def before_request():
pass
@portal.route("/dashboard/")
def dashboard():
return render_template("dashboard.html")// ${ } delimiters keep {{ }} free for Jinja
const app = createApp({
delimiters: config.delimiters,
mixins: [layoutMixin],
data: () => ({ items: [], total: 0 }),
async mounted() {
const { data } = await axios.get('/api/users')
this.items = data.items
this.total = data.total
}
})
registerEnfernoComponents(app)Any agent that reads AGENTS.md or agent skills: Claude Code, Codex, Cursor, Copilot and Gemini CLI. The skill lives in .agents/skills/, the shared location, and Claude Code sees it through .claude/skills.
Not necessarily. Your agent writes the Vue code following the project's patterns, and Enferno works with vanilla JavaScript too. Vue 3 and Vuetify are there when you want a rich component library.
Yes. It's plain Flask with a CLI. "What your agent will run" above lists the five commands to get running.
Absolutely. MIT license gives you complete freedom to build and sell commercial applications without any restrictions or attribution requirements.
Enferno provides production-ready defaults: authentication with 2FA, passkeys and OAuth, Alembic migrations, a Vue 3 admin with no build step, and Docker deployment. Redis, Celery and PostgreSQL are opt-in. It's the boilerplate you'd build anyway, already battle-tested.
Full source, commercial use, no attribution required. Point your agent at it and ship.