README
¶
AICoF - Artificial Intelligence Coding Framework
Build smarter, faster, and more scalable software Cross-tool compatible • Opinionated guardrails • Tech-stack agnostic • Token-optimized
What's New in v1.7.0
- Rebranded to AICoF - Artificial Intelligence Coding Framework with clearer identity
- 33 Framework Guides - Comprehensive framework-specific guidance across 11 language families
- 21 Language Guides - All major programming languages covered
- AGENTS.md Compatible - Works with Claude Code, Cursor, Codex, Copilot, and 20+ other AI tools
Quick Start (60 Seconds)
Option 1: Using the CLI (Recommended)
# Install CLI
brew tap ar4mirez/tap
brew install aicof
# Or via curl
curl -sSL https://raw.githubusercontent.com/ar4mirez/aicof/main/install.sh | sh
# Initialize your project
aicof init my-project
cd my-project
# Explore available components
aicof search react # Find components by keyword
aicof info framework react # View component details
aicof list --available # List all components
# Start coding with AI - guardrails apply automatically!
Option 2: Manual Copy
# 1. Copy template files to your project
cp -r /path/to/aicof/template/{CLAUDE.md,AI_INSTRUCTIONS.md,.agent} ./
# 2. (Optional) For cross-tool compatibility
ln -s CLAUDE.md AGENTS.md
# 3. Start coding with AI - guardrails apply automatically!
The system works immediately:
- AI loads CLAUDE.md automatically (500 lines of guardrails + operations)
- Language guides auto-load based on file extensions
- Workflows available when you need them
- Progressive - starts minimal, grows with your project
Read the full Quick Start Guide
Cross-Tool Compatibility (AGENTS.md)
This system follows the AGENTS.md standard - the universal format for AI coding assistants adopted by 20,000+ repositories.
How It Works
| Tool | Primary File | Fallback |
|---|---|---|
| Claude Code | CLAUDE.md | AGENTS.md |
| Cursor | AGENTS.md | CLAUDE.md |
| OpenAI Codex | AGENTS.md | CLAUDE.md |
| GitHub Copilot | AGENTS.md | CLAUDE.md |
| Google Jules | AGENTS.md | CLAUDE.md |
Setup for Cross-Tool Teams
Option 1: Symlink (recommended)
ln -s CLAUDE.md AGENTS.md
Option 2: Generate standalone AGENTS.md
@.agent/workflows/generate-agents-md.md
Why both files?
CLAUDE.md= Full methodology (guardrails + 4D + workflows)AGENTS.md= Operations only (commands, boundaries, style)
What Is AICoF?
AICoF (Artificial Intelligence Coding Framework) is an opinionated AI development framework designed for professional software teams.
Key Features
| Feature | Description |
|---|---|
| 35+ Guardrails | Testable rules, not vague suggestions |
| 21 Language Guides | All major languages with auto-loading support |
| 33 Framework Guides | Framework-specific patterns and best practices |
| 13 Workflows | PRD, tasks, init, troubleshooting, code-review, and more |
| 3 Modes | ATOMIC/FEATURE/COMPLEX (scales from bugs to architecture) |
| 4D Methodology | Deconstruct → Diagnose → Develop → Deliver |
| Cross-Tool | Works with any AI coding assistant |
Philosophy
Small, validated changes. Quality enforced. Documentation grows organically.
CLI Commands
The aicof CLI manages framework installation, updates, and component discovery.
Core Commands
| Command | Description | Example |
|---|---|---|
init [project] |
Initialize AICoF in a project | aicof init my-app |
update |
Update to latest framework version | aicof update |
doctor |
Check installation health | aicof doctor |
version |
Show CLI and framework versions | aicof version |
Component Management
| Command | Description | Example |
|---|---|---|
add <type> <name> |
Add a component | aicof add framework react |
remove <type> <name> |
Remove a component | aicof remove language rust |
list [--available] |
List installed/available components | aicof list --available |
Type aliases: language (lang, l), framework (fw, f), workflow (wf, w)
Discovery Commands
| Command | Description | Example |
|---|---|---|
search <query> |
Search components by keyword | aicof search api |
info <type> <name> |
Show component details | aicof info fw nextjs |
diff [v1] [v2] |
Compare versions | aicof diff v1.6.0 v1.7.0 |
Configuration
| Command | Description | Example |
|---|---|---|
config list |
Show all config values | aicof config list |
config get <key> |
Get a config value | aicof config get version |
config set <key> <value> |
Set a config value | aicof config set registry https://... |
Valid keys: version, registry, installed.languages, installed.frameworks, installed.workflows
Command Examples
# Initialize and customize
aicof init my-project
aicof add lang typescript
aicof add fw react nextjs
aicof add wf code-review security-audit
# Discover components
aicof search python # Fuzzy search across all types
aicof search --type fw api # Search only frameworks
aicof info lang go --preview 20 # Preview first 20 lines
# Manage updates
aicof diff # Compare installed vs latest
aicof diff v1.6.0 v1.7.0 # Compare specific versions
aicof update # Apply updates
# Troubleshoot
aicof doctor # Check installation health
aicof config list # View current configuration
Professional Prompt Examples
Bug Fixes (ATOMIC Mode)
Fix the null pointer exception in UserService.getProfile() when user.email is undefined
The checkout button is disabled after form validation passes - debug and fix
Memory leak in useWebSocket hook - component doesn't cleanup on unmount
Feature Development (FEATURE Mode)
Add pagination to the /api/users endpoint with cursor-based navigation
Implement email verification flow: send verification link, validate token, update user status
Create a reusable DataTable component with sorting, filtering, and column resizing
Complex Features (COMPLEX Mode)
@.agent/workflows/create-prd.md
Build a real-time notification system with WebSocket support, push notifications,
and user preference management
@.agent/workflows/create-prd.md
Implement multi-tenant architecture: tenant isolation, data partitioning,
tenant-specific configurations
@.agent/workflows/create-prd.md
Add comprehensive audit logging: user actions, data changes, security events,
with retention policies and export functionality
Code Review & Analysis
Review the authentication module for security vulnerabilities and suggest improvements
Analyze the database query patterns in OrderService and identify N+1 query issues
Evaluate the error handling strategy in the API layer and propose a consistent approach
Refactoring
Refactor UserController to use dependency injection and separate business logic into a service layer
Extract common validation logic from all form components into a reusable validation hook
Migrate the codebase from callbacks to async/await while maintaining backward compatibility
Architecture & Planning
@.agent/workflows/create-prd.md
Design a caching strategy for the product catalog: cache invalidation,
distributed caching, cache warming
Analyze the current monolith and propose a microservices decomposition plan
with clear service boundaries
Create a database migration strategy for adding multi-region support
without downtime
Debugging & Troubleshooting
@.agent/workflows/troubleshooting.md
Production error: "Connection pool exhausted" appearing intermittently under load
Performance degradation: API response times increased from 50ms to 500ms
after last deployment
Investigate why the CI pipeline is flaky - tests pass locally but fail
randomly in GitHub Actions
Testing
Generate comprehensive unit tests for the PaymentService with edge cases
for failed transactions, refunds, and currency conversion
Create integration tests for the OAuth flow covering Google, GitHub,
and email/password authentication
Add E2E tests for the checkout flow using Playwright, including
error scenarios and payment failures
Documentation
Generate API documentation for the /api/v2/orders endpoint including
request/response examples and error codes
Create a technical design document for the upcoming search feature
with architecture diagrams
DevOps & Infrastructure
Create a Dockerfile for the Node.js API with multi-stage build,
non-root user, and health checks
Write GitHub Actions workflow for: lint, test, build, deploy to staging
on PR, deploy to production on merge
How It Works
The 3 Modes
ATOMIC (<5 files, clear scope)
"Fix the login button alignment"
- Direct implementation
- Quick validation
- One commit
- Examples: Bug fixes, styling, simple features
FEATURE (5-10 files)
"Add user profile editing with avatar upload"
- Break into 3-5 subtasks
- Implement sequentially
- Integration testing
- Examples: New component, API endpoint, refactoring
COMPLEX (>10 files, new subsystem)
@.agent/workflows/create-prd.md
"Build user authentication with OAuth"
- Create PRD (Product Requirements Document)
- Generate task breakdown
- Step-by-step implementation
- Examples: Authentication, payments, analytics
AI auto-detects which mode to use.
The Guardrails (35+ Rules)
Code Quality:
- Functions ≤50 lines
- Files ≤300 lines
- Complexity ≤10 per function
- All exports have types/docs
Security (CRITICAL):
- All inputs validated
- Parameterized queries only
- No secrets in code
- Dependencies checked for vulnerabilities + licenses
Testing (CRITICAL):
-
80% coverage for business logic
-
60% overall coverage
- Tests for all public APIs
- Regression tests for bugs
Git:
- Conventional commits (
feat:,fix:, etc.) - One logical change per commit
- All tests pass before push
- PRs required (no direct commits to main)
See all guardrails in CLAUDE.md
The 4D Methodology
┌─────────────┐ ┌─────────────┐ ┌─────────────┐ ┌─────────────┐
│ DECONSTRUCT │ ──▶ │ DIAGNOSE │ ──▶ │ DEVELOP │ ──▶ │ DELIVER │
│ │ │ │ │ │ │ │
│ Break down │ │ Identify │ │ Implement │ │ Validate │
│ the task │ │ risks & │ │ with tests │ │ & commit │
│ │ │ dependencies│ │ │ │ │
└─────────────┘ └─────────────┘ └─────────────┘ └─────────────┘
Use Cases
Starting a New Project
@.agent/workflows/initialize-project.md
"Initialize a new TypeScript API with Express, PostgreSQL, and Jest"
AI will:
- Ask clarifying questions (architecture, deployment target)
- Create directory structure
- Generate config files (tsconfig, package.json, etc.)
- Create
.agent/project.mddocumenting decisions - Set up testing framework
Onboarding to Existing Project
@.agent/workflows/initialize-project.md
"This is an existing project - analyze the codebase and document patterns"
AI will:
- Scan tech stack (package.json, requirements.txt, etc.)
- Analyze code patterns and conventions
- Review recent commits
- Create
.agent/project.mdwith findings - Extract patterns to
.agent/patterns.md
Building Complex Features
@.agent/workflows/create-prd.md
"Build a subscription billing system with Stripe integration"
AI will:
- Ask clarifying questions (plans, trial periods, webhooks)
- Create comprehensive PRD
- Generate task breakdown (20-50 subtasks)
- Implement step-by-step with verification
- Update documentation
Cross-Tool Team Workflow
@.agent/workflows/generate-agents-md.md
"Generate AGENTS.md for our team using Cursor and Claude Code"
AI will:
- Extract Operations section from CLAUDE.md
- Generate universal AGENTS.md
- Ensure both tools work consistently
Documentation
| Document | Purpose | When to Read |
|---|---|---|
| AI_INSTRUCTIONS.md | Quick Start Guide | Read this first |
| CLAUDE.md | Core guardrails & methodology | AI loads automatically |
| .agent/README.md | .agent/ folder structure | When customizing |
Language Guides (Auto-Load)
| Language | Files | Guide |
|---|---|---|
| TypeScript/JavaScript | .ts, .tsx, .js, .jsx |
typescript.md |
| Python | .py |
python.md |
| Go | .go |
go.md |
| Rust | .rs |
rust.md |
| Kotlin | .kt, .kts |
kotlin.md |
| Java | .java |
java.md |
| C# | .cs |
csharp.md |
| PHP | .php |
php.md |
| Swift | .swift |
swift.md |
| C/C++ | .c, .cpp, .h, .hpp |
cpp.md |
| Ruby | .rb |
ruby.md |
| SQL | .sql |
sql.md |
| Shell/Bash | .sh, .bash |
shell.md |
| R | .r, .R |
r.md |
| Dart/Flutter | .dart |
dart.md |
| HTML/CSS | .html, .css, .scss |
html-css.md |
| Lua | .lua |
lua.md |
| Assembly | .asm, .s |
assembly.md |
| CUDA | .cu, .cuh |
cuda.md |
| Solidity | .sol |
solidity.md |
| Zig | .zig |
zig.md |
Framework Guides (On-Demand)
| Language | Frameworks |
|---|---|
| TypeScript/JS | React, Next.js, Express |
| Python | Django, FastAPI, Flask |
| Go | Gin, Echo, Fiber |
| Rust | Axum, Actix-web, Rocket |
| Kotlin | Spring Boot, Ktor, Android Compose |
| Java | Spring Boot, Quarkus, Micronaut |
| C# | ASP.NET Core, Blazor, Unity |
| PHP | Laravel, Symfony, WordPress |
| Swift | SwiftUI, UIKit, Vapor |
| Ruby | Rails, Sinatra, Hanami |
| Dart | Flutter, Shelf, Dart Frog |
Workflows (On-Demand)
| Workflow | Purpose |
|---|---|
| Initialize Project | Setup new/existing projects |
| Create PRD | Plan complex features |
| Generate Tasks | Break PRDs into tasks |
| Code Review | Pre-commit quality review |
| Security Audit | Security assessment |
| Testing Strategy | Test planning & coverage |
| Refactoring | Technical debt remediation |
| Dependency Update | Safe dependency updates |
| Troubleshooting | Debug systematically |
| Cleanup Project | Prune unused guides |
| Document Work | Capture patterns & decisions |
| Update Framework | Update AICoF safely |
| Generate AGENTS.md | Cross-tool compatibility |
Repository Structure
aicof/
├── template/ # Distributable template files
│ ├── CLAUDE.md # Main AI instructions (copy to your project)
│ ├── AI_INSTRUCTIONS.md # Quick start guide
│ └── .agent/ # AI context directory
│ ├── language-guides/ # 21 language-specific guides
│ ├── framework-guides/ # 33 framework-specific guides
│ └── workflows/ # 13 structured workflows
├── cmd/aicof/ # CLI entry point
├── internal/ # CLI implementation (commands, core, ui)
├── docs/ # Documentation website source
└── .agent/ # This project's own AI context
├── tasks/ # PRDs and task lists
└── memory/ # Decision logs
Why this structure?
template/contains files distributed to users via the CLIpackages/cli/contains the CLI tool that manages installations.agent/at root is for developing AICoF itself (dogfooding)
System Stats
| Metric | Value |
|---|---|
| Version | 1.7.0 |
| Status | Production Ready |
| AGENTS.md | Compatible |
| Total Files | 67 markdown files |
| CLAUDE.md | ~500 lines |
| Language Guides | 21 (all major programming languages) |
| Framework Guides | 33 (across 11 language families) |
| Workflows | 13 (PRD, tasks, init, troubleshoot, code-review, etc.) |
| Guardrails | 35+ testable rules |
Learning Path
Week 1: Learn the Basics
- Install AICoF via CLI or copy template files to your project
- Write 5 features using ATOMIC mode
- Review guardrails in CLAUDE.md
- Check which language guide applies to you
Week 2: Try Complex Features
- Use PRD workflow for a medium feature
- Generate task breakdown
- Implement step-by-step
- Notice how
.agent/project.mdgrows
Week 3: Customize & Extend
- Add project-specific patterns to
.agent/patterns.md - Create first decision log in
.agent/memory/ - (Multi-tool) Set up AGENTS.md for your team
- Experiment with different modes
FAQ
Q: Do I need to use Claude Code?
A: No! While designed for Claude Code, the system works with any AI coding assistant. Use the AGENTS.md symlink for other tools like Cursor or Codex.
Q: Which file should I edit - CLAUDE.md or AGENTS.md?
A: Edit CLAUDE.md (the source of truth). If using symlink, AGENTS.md updates automatically. If using standalone AGENTS.md, regenerate it after CLAUDE.md changes.
Q: My language isn't covered - what do I do?
A: The core guardrails in CLAUDE.md are language-agnostic (90% still applies). Consider contributing a new language guide!
Q: Is the PRD workflow required?
A: No! Only for complex features (>10 files, new subsystems). Most work is ATOMIC or FEATURE mode.
Q: Can I customize the guardrails?
A: Yes! Edit CLAUDE.md for your team. Common customizations:
- File length limits (300 → 500)
- Coverage targets (80% → 90%)
- Commit format (add your types)
Contributing
Contributions welcome! Ideas:
- Additional language guides (Scala, Elixir, Haskell, OCaml, Julia)
- Additional framework guides for existing languages
- Integration examples with other AI tools
- Real-world case studies
To contribute:
- Fork this repository
- Create a feature branch
- Make your changes following the guardrails
- Submit a pull request
License
MIT License - See LICENSE file for details.
Acknowledgments
Built with:
- Claude Code - Anthropic's AI coding assistant
- AGENTS.md Standard - Universal AI agent instructions
- 4D Methodology - Systematic problem-solving approach
- Community feedback - Continuous improvement
Support
- Documentation: AI_INSTRUCTIONS.md
- Issues: GitHub Issues
- Discussions: GitHub Discussions
Happy coding with AI!
Works with Claude Code, Cursor, Codex, Copilot, and any AI assistant that reads AGENTS.md