OSpec v1.0

Build Products That Deliver Outcomes

The specification format for AI coding agents to build products with proven tech stacks.

OSpec is an outcome-driven specification that defines what product to build and which tech stack to use, allowing AI coding agents to scaffold, implement, and deploy projects with opinionated, battle-tested technology choices.

Why OSpec?

🤖 End the Tech Stack Decision Paralysis

You and your AI coding agents face the same endless choices: “React or Vue? Next.js or Remix? Svelte or React? Supabase or Firebase? npm or pnpm? Which versions actually work together?”

This decision paralysis affects both humans and AI agents. While you debate frameworks, competitors ship products.

OSpec solves this with shared, battle-tested recipes—like having 1000+ developer teams tell you exactly what tech combination works for your use case.

⚡ Skip the Setup, Start Building

Like rails new or create-react-app, but for any stack and any AI agent. Drop in an OSpec file and your preferred coding agent scaffolds the entire project with proven tools, versions, and configurations.

The community shares the setup knowledge. You focus on building your product.

🎯 Outcome-Driven Development

Specify what you want (e-commerce site, CLI tool, API service) instead of how to build it. Multiple AI agents can work together—one for setup, another for deployment, another for testing.

🔧 Multi-Agent Optimization

Different coding agents excel at different tasks. OSpec specs can specify:

Quick Start

Create an outcome.yaml file that tells your AI coding agents exactly what to build:

ospec_version: "1.0.0"
id: "modern-shop"
name: "Modern E-commerce Platform" 
description: "High-performance shop with SSR and edge deployment"
outcome_type: "web-app"

# 2025 battle-tested stack - proven by 500+ projects
technology_stack:
  meta_framework: "Next.js@15"      # 78% of React projects use Next.js
  ui_library: "React@18"            # Still #1 with 39.5% popularity
  styling: "TailwindCSS@4"          # v4 with native CSS integration
  database: "Supabase"              # PostgreSQL + real-time + auth
  payments: "Stripe@13"             # Latest API with enhanced UX
  deployment: "Vercel"              # Edge runtime + ISR
  package_manager: "pnpm@9"         # 3x faster than npm, space efficient

# Multi-agent workflow
agents:
  primary: "claude-code"            # Main development agent
  deployment: "deployment-specialist" # Handles CI/CD setup
  testing: "test-engineer"          # E2E and unit test coverage
  
# Sub-agents with specific expertise (auto-created in .claude/agents/)
sub_agents:
  - name: "ecommerce-specialist"
    description: "Expert in payment flows, cart logic, and product management"
    focus: ["stripe-integration", "inventory-management", "order-processing"]
  
  - name: "performance-auditor" 
    description: "Optimizes Core Web Vitals, bundle size, and loading speed"
    focus: ["lighthouse-scores", "bundle-analysis", "caching-strategy"]

# Embedded automation scripts
scripts:
  setup: |
    #!/bin/bash
    echo "🚀 Setting up Modern Shop with Next.js 15..."
    pnpm create next-app@latest . --typescript --tailwind --app
    pnpm add @supabase/supabase-js stripe lucide-react
    
  deploy: |
    #!/bin/bash
    echo "🌍 Deploying to Vercel with edge optimizations..."
    vercel --prod --regions=fra1,sfo1,hnd1

# How to verify it works
acceptance:
  performance:
    - "Lighthouse score > 95 for all metrics"
    - "First Contentful Paint < 1.2s"
    - "Bundle size < 200KB gzipped"
  functionality:
    - "User can browse 1000+ products with search"
    - "Cart persists across sessions"
    - "Stripe checkout completes in test mode"
    - "Order confirmation emails are sent"

Multi-Agent Workflow:

  1. Claude Code scaffolds the Next.js structure and core components
  2. Deployment Specialist sets up Vercel config and CI/CD pipeline
  3. Test Engineer creates Playwright tests for critical user flows
  4. Performance Auditor optimizes bundle and implements caching
  5. All agents validate against your acceptance criteria

Alternative 2025 Stacks:

# Performance-First Stack (Svelte)
technology_stack:
  meta_framework: "SvelteKit@2"     # 180% growth, 70% smaller bundles
  styling: "TailwindCSS@4"
  database: "PlanetScale" 
  deployment: "Cloudflare Pages"    # Edge-first with global distribution

# Full-Stack TypeScript (Remix)  
technology_stack:
  meta_framework: "Remix@2"         # Server-first, Web Standards
  runtime: "Bun@1.1"               # 3x faster than Node.js
  database: "Prisma + PostgreSQL"
  deployment: "Railway"

Get Started →

Key Features

📋 Comprehensive Project Recipes

YAML files that specify outcomes, tech stacks, agent workflows, and acceptance criteria—like a package.json for entire multi-agent projects.

🏗️ 2025 Battle-Tested Stacks

Modern combinations that actually work: Next.js 15 + Supabase + Vercel, SvelteKit + PlanetScale + Cloudflare, Remix + Bun + Railway.

🤖 Multi-Agent Orchestration

🔧 Embedded Automation

🌐 Community-Driven Knowledge

Share successful recipes, discover proven combinations, skip the research phase. 2000+ hours of setup experience, distilled into reusable specs.

Learn More

📚 Read the Spec

Complete OSpec v1.0 specification with schema, lifecycle, and examples.

🚀 Guides

Step-by-step tutorials for writing OSpecs and building compatible agents.

💡 Examples

Real-world OSpec examples from simple websites to ML pipelines.

📖 Cookbook

Common patterns, troubleshooting tips, and best practices.

Community

OSpec is an open specification format for the AI coding era. Help us curate the best tech stacks and improve AI agent productivity.