Backend Architecture Intelligence for AI Coding Assistants
A senior backend architect embedded inside your AI, installable via npm.
IronBackend is a backend-focused architecture knowledge base designed to inject production-grade architectural intelligence into AI coding assistants like Cursor, Claude, GitHub Copilot, and ChatGPT.
- ❌ A framework or library
- ❌ A runtime dependency
- ❌ A code generator
- ❌ A tutorial or course
- ✅ A backend architecture knowledge base
- ✅ A style & stack selector for AI assistants
- ✅ A rule-driven prompt injection system
- ✅ Distributed via npm CLI
Modern AI coding assistants lack structured backend architectural knowledge:
| Problem | IronBackend Solution |
|---|---|
| Inconsistent architecture advice | 10 production-grade architecture styles |
| Framework-specific suggestions | Stack-agnostic patterns with stack-specific conventions |
| No design principle enforcement | 50+ enforceable design rules |
| Ad-hoc security patterns | Security & reliability playbooks |
# Install globally
npm install -g @ironbackend/cli
# Or use with npx
npx @ironbackend/cli init cursor# Initialize for a specific AI tool
ironbackend init cursor # Cursor IDE
ironbackend init claude # Claude Code
ironbackend init copilot # GitHub Copilot
ironbackend init antigravity # Antigravity
# Interactive mode (select tool from menu)
ironbackend init
# With style and stack pre-selected
ironbackend init cursor --style clean-monolith --stack node-nestjs# List available options
ironbackend list
# Select architecture style
ironbackend select style hexagonal
# Select tech stack
ironbackend select stack python-fastapi
# Re-export prompts after changes
ironbackend export promptsIronBackend provides 10 production-grade architecture styles:
| Style | Best For |
|---|---|
| Clean Monolith | Small teams, startups, single deployable |
| Modular Monolith | Growing teams, bounded contexts |
| Hexagonal | Long-lived systems, test-driven development |
| Event-Driven | High throughput, eventual consistency |
| CQRS | Read-heavy workloads, complex queries |
| Microservices (Sync) | Multiple teams, independent scaling |
| Microservices (Async) | Decoupled services, message queues |
| Serverless | Variable traffic, managed infrastructure |
| Read-Heavy API | High read:write ratio, aggressive caching |
| Automation/Bot | Scheduled jobs, integrations, workflows |
Each style includes:
- When to use / when NOT to use
- Core principles
- Folder structure
- Common pitfalls
- AI instruction snippets
Pre-configured stacks with conventions:
| Stack | Components |
|---|---|
| Node.js + NestJS | TypeScript, Prisma, BullMQ, Jest |
| Java + Spring Boot | Spring Data JPA, RabbitMQ, JUnit 5 |
| .NET + ASP.NET Core | EF Core, MassTransit, xUnit |
| Python + FastAPI | SQLAlchemy, Celery, pytest |
50+ enforceable rules across 8 categories:
- API Design - RESTful conventions, versioning, pagination
- Domain Modeling - Entities, value objects, aggregates
- Error Handling - Typed exceptions, error codes, logging
- Transactions - Scope, timeouts, saga pattern
- Data Access - Repository pattern, N+1 prevention
- Naming - Consistent conventions across languages
- Validation - Boundary validation, schema enforcement
- Async/Concurrency - Idempotency, dead letters, timeouts
ironbackend init <tool> # Initialize for AI tool
ironbackend init # Interactive tool selection
ironbackend select style <name> # Select architecture style
ironbackend select stack <name> # Select tech stack
ironbackend export prompts # Export AI prompts
ironbackend doctor # Validate setup
ironbackend list # List styles & stacks
ironbackend info # Show current config| Command | AI Tool | Output Path |
|---|---|---|
init claude |
Claude Code | CLAUDE.md |
init cursor |
Cursor IDE | .cursor/rules/ironbackend.mdc |
init windsurf |
Windsurf | .windsurfrules |
init antigravity |
Antigravity | .gemini/settings/prompts.md |
init copilot |
GitHub Copilot | .github/copilot-instructions.md |
init kiro |
Kiro | .kiro/rules.md |
init codex |
Codex | AGENTS.md |
init gemini |
Gemini CLI | GEMINI.md |
init trae |
Trae | .trae/rules.md |
Each tool gets a correctly formatted configuration file at the expected location.
After setup, your AI assistant receives context like:
# IronBackend System Prompt
You are a senior backend engineer with 10+ years of experience
in Hexagonal Architecture using Python/FastAPI.
## Core Principles
1. Domain and application core have no external dependencies
2. Ports define what the application needs or provides
3. Adapters implement ports for specific technologies
## Enforced Rules
[ERROR] API-001: All endpoints must have explicit schemas
[ERROR] DOM-005: Domain logic must not depend on framework code
[ERROR] ERR-003: Never expose stack traces in production
## When Generating Code
- Controllers handle HTTP only, delegate to services
- All external dependencies injected via interfaces
- Include proper error handling per ERR-* rules| Package | npm | Description |
|---|---|---|
@ironbackend/cli |
CLI tool | |
@ironbackend/core |
Styles, rules, stacks | |
@ironbackend/prompts |
Prompt templates |
IronBackend embodies:
- Opinionated but configurable - Strong defaults, optional overrides
- Deterministic - Same inputs produce same outputs
- AI-readable - Optimized for LLM context windows
- Production-grade - Patterns from real-world systems
- Framework-agnostic - Architectural patterns, not framework tutorials
Contributions welcome! See CONTRIBUTING.md for guidelines.
Areas to contribute:
- New architecture styles
- Additional tech stacks
- More design rules
- Integration improvements
MIT License - see LICENSE
IronBackend — A senior backend architect in your AI, installable via npm.
