Start here — guest try-out (landing-first)
Type what you need. We ship a page you can preview.
Guests: no login. This path optimizes for a single credible page. For multi-tenant apps, backends, integrations, and long-lived products, open Admin — same agents and gates, richer delivery profile.
One click queues a page you can track · sandbox after QA · optional checkout — or switch to Admin when you need a full application instead.
Launch-ready pages and real apps — from the same brief
Try the line above for a fast, shareable marketing page you can preview after QA — ideal for promos, waitlists, and “show me something real” moments. When scope grows into APIs, data, auth, and multi-screen products, the same agent pipeline runs deeper builds from Admin or Director. Sandbox preview and optional on-chain checkout apply to both paths.
Below: live examples from brochure landings to full-stack listings. The phrase box is intentionally light; Admin is where complex work queues — one engine, not a toy “lite” stack.
Built for speed and substance
Landings are the default guest deliverable; autonomous and admin-queued ideas often become full products — one pipeline, different delivery_profile, identical gates.
One phrase → presentable page
Your sentence becomes the stakeholder brief through the same stages as autonomous builds — HTML/CSS/JS you preview in the sandbox; gates reject hollow stubs.
Specialized agents
Twelve pipeline roles (Analyst, PM, Methodologist, Architect, Designer/UX, Developer, QA, Security, DevOps, Marketing, Sales, Evolution) — each step bounded so outputs stay maintainable.
Quality & security gates
Demo checks, headless browser smoke, optional marketplace rules — rework loops until the product is show-ready.
Same runway as autonomous
Research → spec → architecture → code → QA → security → DevOps → marketing → sales → evolution. Autonomous seeds ideas from the market; on-demand seeds from your phrase — no second-class pipeline.
Director AI
Meta-agent reviews pipeline health on a schedule and steers autonomous improvements.
Crypto-ready storefront
Affordable one-shot landing price (about $5 USDT when agents omit list pricing), multi-chain checkout — buyers pay on-chain, you ship files.
One pipeline, two front doors
Autonomous mode feeds market research and generated ideas; on-demand uses your phrase as the brief. Same agent path — spec, build, QA, and beyond.
Research
Spec
Marketing
Architect
Design
Develop
QA
Launch
Product experience
Designer layer — modern UI by default
Before code ships, the Architect emits a structured `ui_experience` brief: tokens, typography, motion, and a signature visual moment. The Developer treats it as binding for browser deliverables — so landings read as intentional product design, not generic AI gray boxes.
Runtime topology
Architecture at a glance
One control plane: web tier, background workers, routed models, and durable workspace — shown as a live orbit around the agent fleet.
Agents
PM · Architect · Dev · QA · Sec · Ops · Mkt · Sales · Evolution
Next.js
Storefront
FastAPI
Public & admin API
Pipeline worker
Quality gates
Director AI
Signals & reports
LLM router
Multi-provider
Data plane
SQLite · artifacts
Marketplace examples
Browse separately: brochure landings (hero generator) vs. full products (admin / autonomous pipeline). All listings passed quality gates.
Quality Trust Block
Live benchmark-based quality metrics from production pipeline.
Loading metrics…
FAQ
What does AI-Factory generate?
Primarily share-ready marketing landings from the public phrase box; the same pipeline also produces full_software builds (APIs, persistence, compose-friendly repos) when work is queued from Admin or Director.
Can I run multiple ideas at once?
Yes, batch mode supports up to 10 ideas with progress tracking and retry support.
How does quality control work?
Quality uses benchmark pass-rate metrics, release gates, and continuous pipeline checks.
Do you support paid upgrades?
Yes. Free tier limits are visible in account, and paid plans upgrade through Stripe checkout.
Ready to ship your next page or product?
Self-host the factory, connect LLM keys, then use the phrase box for a landing or Admin for a full build — same agents, same quality bar.