AI Development Methodology

Build Production-Grade
Software with AI

An AI development methodology that takes you from research to shipped code — with enterprise-grade quality gates at every step.

~

What is Raftkit?

A methodology and toolkit for building production-grade software with AI — from idea to shipped code.

📐

Methodology

A 6-phase structured workflow that takes you from research to production code. Not a tool — a way of building software.

🔍

Quality Gates

Two-stage review with up to 12 specialized AI reviewers per task. P1/P2/P3 severity system blocks bad code from shipping.

Zero Config

One command installs everything. Auto-detects your stack, package manager, and project type. Works with any codebase.

The 6-Phase Workflow

From research to compounding knowledge — a structured methodology for building with AI.

See It In Action

One command to set up your entire development workflow.

~/my-project

What Gets Built

Every phase produces concrete, versioned artifacts — not just code.

Project Structure
docs/products/my-saas/
├── prd.md
├── features.md
├── architecture/
│   ├── tech-preferences.md
│   ├── db-schema.md
│   ├── tech-stack.md
│   ├── system-diagram.md
│   ├── api-contracts.yaml
│   └── design-system.md
├── tasks/
│   ├── index.md
│   ├── task-001.md
│   ├── task-002.md
│   └── ...
├── reviews/
│   ├── task-001-review.md
│   └── ...
├── learnings/
│   ├── index.yaml
│   ├── task-001-learning.md
│   └── compound-summary.md
└── status.md
Research

PRD

Competitive analysis, personas, market sizing

prd.md
Architect

Architecture

DB schema, API contracts, system diagram, design tokens

architecture/
Plan

Micro-tasks

RICE-prioritized, 2-5 min each, EARS acceptance criteria

tasks/
Build

Code

TDD per task, fresh subagent, auto-committed

src/
Review

Reviews

12-dimension quality reports per task

reviews/
Compound

Learnings

Compound knowledge that evolves your project docs

learnings/

Detects Your Stack. Installs Matching Skills.

Raftkit auto-detects your tech stack and installs matching AI skills — 28 and counting.

Frontend

React 19Next.jsVueAngularSvelteTailwind CSSFramer Motion

Backend

Node.jsExpressHonoNestJSFastAPIDjango

Database

PostgreSQLDrizzle ORMPrismaMongoDBRedis

Quality

Code QualitySEOAccessibilityPerformanceSecurity

Specialized

GraphQLWebSocketDockerAWSVercel

Quality You Can Trust

Every micro-task gets reviewed by up to 12 specialized AI agents before commit.

🛡️Security
Code Quality
Performance
🏗️Architecture
🧪Test Coverage
🔬Overbuilt
🚨Error Handling
Accessibility
📡API Contracts
🗄️Database
🔌Third Party
📱Responsive

Two-Stage Review

Stage 1
Spec compliance — does the implementation match the task requirements?
passes? →
Stage 2
12-dimension review — conditional dispatch to relevant reviewers based on task type.

Severity System

P1
Blocks commitMust be fixed before code can be merged
P2
Should fixImportant issues that should be addressed
P3
LoggedMinor suggestions tracked for future improvement

Get Started in 5 Steps

From zero to autonomous development in minutes.

1

Install

One command sets up Git hooks, commit conventions, and AI skills

npx @raftlabs/raftkit init
2

Research

Deep-dive into your product space with competitive & user research

/raftkit:research my-product
3

Architect

Design your system architecture, DB schema, and API contracts

/raftkit:architect my-product
4

Plan

Break features into micro-tasks with acceptance criteria

/raftkit:plan my-product
5

Build

Autonomous TDD build loop with quality gates at every step

/raftkit:build my-product
MIT License

Open Source

Built by RaftLabs for the developer community. Free to use, modify, and distribute.