rook-framework 0.2.0

Agent-native SaaS infrastructure. Build, modify, and operate production backends safely with AI agents.

Rook

Agent-native SaaS infrastructure. Build, modify, and operate production backends safely with AI agents.

Zero to running multi-tenant backend in 15 minutes — generated by agents via MCP, scoped by capability tokens, checkpointed for safe rollback.

pip install rook-framework
rook init my-app
rook resource generate Invoice --fields "client_id:fk(Client) status:enum(draft,sent,paid) amount:int"
rook dev

What you get in one command

rook init generates a complete production backend:

• Multi-tenant with RLS — row-level security at the database level, not just application filtering
• FastAPI + SQLAlchemy 2.0 + PostgreSQL — async, typed, production-ready
• Auth (JWT + Clerk), billing (Stripe), email (Resend), webhooks (HMAC + retry)
• Background jobs (ARQ), rate limiting, idempotency, feature flags
• CI pipeline — ruff, mypy, migration safety, schema drift, N+1 detection, security audit
• Deploy pipeline — snapshot → migrate → deploy → health gate → auto-rollback
• CLAUDE.md — AI agent context for every generated project
• MCP server — 30+ tools for AI agents to build, deploy, and operate your backend

Code generation

rook resource generate Order --fields "customer_id:fk(Customer) status:enum(pending,paid) total:decimal"

One command creates 7 files: model (with FK relationships + enum binding), schemas, service, controller, tests, factory, and migration (with RLS policies). Routes auto-registered. Migration auto-chained.

Production patterns included

Pattern	What it does
Row-Level Security	Database-enforced tenant isolation
TxnManager	Auto-commit, auto-rollback, RLS via SET LOCAL
Outbox pattern	Reliable webhook delivery
Expand/contract migrations	Zero-downtime schema changes
N+1 detection	Per-request query counting with warnings
Schema drift detection	rook db doctor catches model/DB mismatches
Migration linting	rook db lint catches dangerous patterns
Idempotency middleware	Dedup POST/PUT/PATCH by Idempotency-Key
Health-gated deploys	Auto-rollback if deploy fails health check

Agent-native

Rook is designed from the ground up to be operated by AI agents:

rook mcp serve

Exposes 30+ tools via Model Context Protocol. An AI agent can:

• Generate resources and wire them up
• Run migrations safely
• Deploy with checkpoints and rollback
• Detect schema drift and fix it
• Monitor query performance

Stack

Layer	Default
Backend	FastAPI + SQLAlchemy 2.0 (async)
Database	PostgreSQL + pgvector
Auth	JWT (Clerk optional)
Billing	Stripe
Email	Resend
Jobs	Redis + ARQ
Storage	Cloudflare R2
Monitoring	Sentry + structlog
CI/CD	GitHub Actions (generated)

Documentation

• Getting Started
• CLI Reference
• Roadmap

License

MIT