Developable

Developable reads a Prisma schema and generates a complete, production-ready Express + TypeScript REST API — with security invariants baked structurally into every file, not as prompts you have to remember to follow.

The difference from asking an LLM to write a backend: the security rules are in the Jinja2 templates. The LLM cannot skip them, forget them, or override them. Every generated API enforces auth middleware on write routes, server-side FK injection, ownership checks before updates and deletes, sensitive-field hashing, and ID validation — regardless of what the model does with the rest of the code.

Use it as a Claude Code skill (/developable), an OpenAI Codex skill, or a Python CLI.

---

How it works with Claude Code

Claude Code is Anthropic's AI coding assistant — available as a CLI, a VS Code / JetBrains extension, and at claude.ai/code. When you type /developable in a Claude Code session, it runs the Developable skill: an instruction set that tells Claude exactly how to parse your schema, which files to write, and which security rules are non-negotiable.

No separate server, no API calls from your machine — Claude Code handles everything using its own context window and built-in file tools.

OpenAI Codex works the same way: the skill ships as AGENTS.md instructions that Codex follows during a session.

---

Install

Claude Code

curl -sSL https://raw.githubusercontent.com/developableai/backend/main/install.sh | bash

Restart Claude Code. That's it — /developable is now available.

The command will appear in the / picker with a description. Type /developable and hit Enter — Claude Code handles the rest interactively.

OpenAI Codex

$skill-installer install https://github.com/developableai/backend/tree/main/skills/developable

Or copy skills/developable/SKILL.md into your project's AGENTS.md manually.

Once installed, start a Codex session in your project directory and say: run /developable or just describe what you want — Codex will follow the skill instructions automatically.

Python CLI

pip install developable
export ANTHROPIC_API_KEY=sk-ant-...

---

Usage

Start from a description (no schema needed)

/developable "A task management app with users, projects, and tasks. Users log in with email."

The skill generates a schema.prisma with correct annotations and a rules.yaml with business constraints, shows them to you for review, lets you iterate, then generates the full API once you confirm.

Start from an existing schema

Annotate your schema.prisma:

// @auth_entity
// @llm Users can only access their own tasks
model User {
  id       Int    @id @default(autoincrement())
  email    String @unique
  password String // @llm sensitive
  tasks    Task[]
}

model Task {
  id      Int    @id @default(autoincrement())
  title   String
  done    Boolean @default(false)
  owner   User   @relation(fields: [ownerId], references: [id])
  ownerId Int
}

Then run:

/developable

The skill reads your schema, generates all API files, and writes them into your project. No Python runtime or API key needed for skill use.

Python CLI

developable schema.prisma --out ./my-api

# Skip LLM calls — instant output, placeholder Zod schemas
developable schema.prisma --out ./my-api --no-llm

# Also generate the integration test suite
developable schema.prisma --out ./my-api --tests-out ./tests

# Push to a new GitHub repo and trigger CI
developable schema.prisma --out ./my-api --github

# Re-run after schema changes — skips files you've manually edited
developable schema.prisma --out ./my-api --no-llm

# Force-overwrite everything including user-modified files
developable schema.prisma --out ./my-api --no-llm --force

After generation:

cd my-api
npm install
npx prisma migrate dev --name init
npm run dev
# → http://localhost:3000

Cloud deploy (CLI only):

python deploy.py --out ./my-api --deploy-to aws
python deploy.py --out ./my-api --deploy-to gcp --gcp-project my-project-id
python deploy.py --out ./my-api --deploy-to heroku

---

What gets generated

Per entity

File	Description
src/routes/<entity>.routes.ts	Express Router; JWT middleware applied per method
src/controllers/<entity>.controller.ts	HTTP handlers, ID validation, ownership guards
src/repositories/<entity>.repository.ts	Prisma data-access layer
src/validators/<entity>.validator.ts	Zod schemas generated by Claude for your domain
src/types/<entity>.types.ts	TypeScript input/output interfaces

Shared infrastructure

File	Description
src/auth.ts	JWT authenticate middleware
src/auth.controller.ts	Register + login, JWT signing, bcrypt
src/errors.ts	AppError hierarchy + error-handler middleware
src/pagination.ts	parsePagination + buildPaginatedResponse
src/app.ts	Express app: helmet, cors, morgan, router mounting

DevOps

File	Description
Dockerfile	Multi-stage Node 20 production image
docker-compose.yml	Local stack: PostgreSQL + pgAdmin + API
.github/workflows/ci.yml	GitHub Actions: install → migrate → start → test
.env.example	All required environment variables

---

REST endpoints

Every entity gets five routes:

Method	Path	Auth
GET	/api/{plural}	Optional
GET	/api/{plural}/:id	Optional
POST	/api/{plural}	Required
PUT	/api/{plural}/:id	Required
DELETE	/api/{plural}/:id	Required

Filtering, sorting, and pagination on all list endpoints:

GET /api/tasks?filter[done]=false&sort=createdAt&order=desc&page=1&limit=20

One-to-many relations generate nested routes automatically:

GET  /api/users/:id/tasks
POST /api/users/:id/tasks

---

Security invariants

Every generated API enforces these unconditionally — they live in the templates, not in prompts:

What is enforced
Integer IDs validated before Prisma is touched — rejects floats, alpha, SQL injection suffixes, overflow
String IDs (uuid/cuid PKs) validated — rejects whitespace and oversized strings
Owner FK stripped from request body, injected server-side from the verified JWT
Auth entity self-ownership — users can only update or delete their own record
Ownership check on every update and delete for resources owned by a user
Sensitive fields hashed with bcrypt before any database write
Sensitive fields excluded from JWT payload and all API responses
JWT verified on all write routes and ownership-sensitive reads
Filter fields validated against an allowlist — sensitive or unknown fields return 400

---

Schema annotations

Annotation	Where	Effect
// @auth_entity	Above a model	Marks the login principal; generates auth controller + JWT middleware
// @llm sensitive	On a field	Hashed at rest; excluded from JWT and all responses
// @llm <hint>	Above a model	Free-text hint forwarded to Claude when generating validators

---

Output scale and cost

Real measurements from full project runs with the Python CLI:

Project	Entities	TypeScript	Tests	Cost
E-commerce API	5	~1,200 lines	~2,400 lines (100+ cases)	$0.08
Project management API	6	~1,450 lines	~2,900 lines	$0.10

The Claude Code and Codex skills produce the same output at zero API cost (Claude Code's own context handles generation).

---

Requirements

Skill use (Claude Code / Codex)	No Python or Node required
CLI use	Python 3.11+, Node 18+, PostgreSQL
Anthropic API key	CLI only — not needed for skill use

---

Marketplace status

Distribution	Status
pip install developable	Published — pypi.org/project/developable
Claude Code skill (install.sh)	Available
Claude Code plugin marketplace	Not yet submitted
OpenAI Codex skill registry	Available via GitHub URL above