opencode-autobe

An OpenCode plugin that integrates AutoBE — the AI-powered NestJS + Prisma backend generator — directly into your OpenCode sessions.

What it does

Adds three tools to OpenCode:

| Tool | Description | |------|-------------| | autobe_generate | Generate a complete NestJS + Prisma backend from a natural-language description | | autobe_list_sessions | List recent AutoBE sessions on your playground server | | autobe_get_files | Retrieve generated files from a completed session and write them to your project |

Prerequisites

1. AutoBE playground server — clone autobe and run the playground server
2. AI API key — Anthropic, OpenAI, or any OpenAI-compatible provider

Installation

As an npm plugin (recommended)

Add to your opencode.json:

{
  "plugins": ["@hacr/opencode-autobe"]
}

Local development

Clone this repo and place the project in your working directory. OpenCode will auto-load .opencode/plugins/autobe.ts.

Configuration

Set these environment variables before starting OpenCode:

| Variable | Default | Description | |----------|---------|-------------| | AUTOBE_SERVER_URL | http://localhost:3000 | AutoBE playground server URL | | AUTOBE_API_KEY | — | API key for the AI vendor (overrides other keys) | | ANTHROPIC_API_KEY | — | Anthropic API key (used if AUTOBE_API_KEY not set) | | OPENAI_API_KEY | — | OpenAI API key (used as last fallback) | | AUTOBE_BASE_URL | — | Custom base URL for OpenAI-compatible endpoints | | AUTOBE_MODEL | claude-sonnet-4-20250514 | Default AI model | | AUTOBE_VENDOR_ID | — | Skip vendor creation, use this existing vendor ID |

Usage

Once configured, just describe your backend in natural language:

Generate a backend for a blog platform with posts, comments, tags, and user auth

OpenCode will call autobe_generate and run the full AutoBE pipeline:

1. Requirements analysis — structures your description into a formal spec
2. Database design — creates a type-safe Prisma schema
3. API design — generates an OpenAPI specification
4. Test generation — writes E2E test suites
5. Code generation — produces compilable NestJS implementation

All generated files are written to your current project directory.

AutoBE pipeline

AutoBE uses a waterfall + spiral architecture with compiler-driven validation:

Requirements → Prisma Schema → OpenAPI Spec → E2E Tests → NestJS Implementation
     ↑               ↑               ↑             ↑               ↑
  Analyze         Database        Interface       Test           Realize

Each phase validates against its compiler (Prisma → OpenAPI → TypeScript) before proceeding.

Starting the playground server

# Clone autobe
git clone https://github.com/samchon/autobe
cd autobe

# Install dependencies
pnpm install

# Set up database
cd apps/playground-server
npx prisma db push

# Start the server (default port from .env.local)
pnpm exec ts-node src/executable/server.ts

Set AUTOBE_SERVER_URL to the server's address and port.

License

MIT