Spec4AI
For Developers & AI Engineers

Turn your engineering knowledge into executable specs

Package your best practices, architecture patterns, and domain expertise into specs that other developers and AI agents can follow to produce reliable, validated output.

What is a Spec?

A spec is a bundle of structured files that tell an AI agent exactly what to build, how to validate it, and what “done” looks like. Think of it as an executable blueprint.

Structured, Not Freeform

Every spec follows a defined file structure. Manifest for metadata, QA for requirements, agent instructions for workflow, and checks for validation. No guessing.

Validated, Not Hoped

Specs include checks that run after generation. Contract checks verify the output matches the declared schema. Smoke checks verify the build runs. No more “works on my machine.”

Versioned, Not Stale

Specs use semantic versioning with stable and canary channels. Buyers get updates through the CLI. Security patches are delivered free, even after support expires.

Two Spec Editions

Choose the right level of structure for your project.

Default

AI Spec

6 core files. Lightweight bundle for AI agents, prompt templates, creative projects, MCP servers, and Claude skills.

sai init my-agent --type ai-agent

Includes:

manifest.md, agent.md, qa.md, status.md, security.md, README.md

Advanced

Engineering Spec

22-30+ files with behaviors, security, tests, UI states, ops, and agent-specific configs. For production software.

sai init my-app --type web-app --engineering

Adds:

behaviors/, edge_cases/, security/, tests/, ui/, ops/, agents/

Create a Spec in 4 Steps

1

Initialize

Run sai init to scaffold the right files for your project type. Choose from 47 artifact types across 10 categories.

sai init my-spec --type rest-api
2

Define

Fill in the manifest (metadata), QA (requirements), and agent instructions (workflow). The CLI validates structure as you go.

sai verify ./my-spec
3

Validate

Add checks that verify the AI output is correct. Contract checks, smoke tests, and schema validation. Score must reach 80% to publish.

sai score ./my-spec
4

Publish

Push your spec to the marketplace. Set pricing, choose a channel (stable or canary), and start reaching developers and AI agents.

sai publish

What You Can Spec

47 artifact types across 10 categories. From web apps to AI workflows.

Web Apps

SPA, PWA, E-Commerce

APIs & Services

REST, GraphQL, gRPC

AI Agents

Chatbots, RAG, Voice

Creative AI

Prompts, Workflows, Video

Infrastructure

Terraform, K8s, CI/CD

CLI Tools

CLI, Browser Ext, VS Code

Libraries

NPM, Pip, Crates, SDKs

Mobile

iOS, Android, Cross-Platform

Data & ML

Pipelines, Models, Dashboards

Embedded

Firmware, RTOS, IoT

Earn From Your Expertise

Publish free specs to build reputation, or set a price and earn from every sale.

You Set the Price

Free, one-time purchase, or subscription. You control the pricing. Revenue is paid out through Stripe Connect directly to your bank account.

Buyers Get Updates

When you publish a new version, buyers receive it through the CLI. Security patches are always free. Buyers trust specs that stay maintained.

Completeness Score

Every spec gets a quality score based on completeness: manifest metadata, QA depth, agent instructions, checks coverage, and security documentation.

Ready to create your first spec?

Install the CLI, scaffold your spec, and publish to the marketplace. The full documentation covers everything from artifact types to validation tiers.