← All Deliverables
AaaS Infrastructure · Partner Briefing

AntiGravity
Explained

How the autonomous AI infrastructure behind Agents-as-a-Service works — from the development engine to the self-monitoring supervisor that runs 24/7.

Last updated: 2026-03-19

1. The 30-Second Version

We built an AI platform called AaaS (Agents-as-a-Service) where businesses email tasks to specialist AI agents and get work done — no app required. Behind the visible websites, there's an autonomous infrastructure layer called AntiGravity that builds features, monitors health, and improves the system continuously without human intervention.

Think of it as: the websites are the storefront, and AntiGravity is the factory, warehouse manager, and quality inspector — all running autonomously.

4
Live Domains
12+
Cloud Functions
6
Automated Scans
4,200+
AI Skills in Vault
75
AutoResearch Agents

2. What Users See — The 4 Domains

Each domain serves a specific purpose in the customer journey:

3. The Customer Journey — Fully Automated

From first visit to active user, everything is automated. No human intervention required except for billing setup.

DISCOVERY User finds us via blog (SEO), direct link, or referral ↓ LANDING (aaas.name) Signs up with Google, Email, or Microsoft ↓ CONTEXT ENGINEERING (automated, 30-60 seconds) 1. Scrape their company domain 2. Classify industry, stage, tech stack (Vertex AI) 3. Generate personalized context package 4. Match relevant skills from 4,200+ vault 5. Suggest best-fit agents from our fleet ↓ ONBOARDING (5-step guided wizard) Context → Agents → First Task → Vault → Dashboard ↓ ACTIVATION EMAIL DRIP (7 stages over 7+ days) Day 0: Welcome email Day 0: Context package ready Day 1: Send your first task Day 3: Meet your agent fleet Day 5: Vault skill highlights Day 7: Book a strategy call Weekly: Activity digest ↓ ACTIVE USER Emails tasks to specialist agents Gets work done, tracks results in dashboard

Key insight: The entire flow — from signup to first task — runs without any human involvement. The context engineering pipeline uses Google's Vertex AI (Gemini 2.0 Flash) to understand each business and personalize the experience in under 60 seconds.

4. What Runs Behind the Scenes — AntiGravity

AntiGravity is the development and operations engine. It consists of three components:

4a. The Development Engine (ADE)

A template system for spinning up AI-powered projects. Every AaaS project is born from the same base — ensuring consistent quality, security, and operational DNA across everything we build.

LayerWhat It DoesExample
1. DirectivesOperating procedures (SOPs)"Always validate inputs before processing"
2. ContextProject-specific memoryBrand guidelines, sprint goals, KPIs
3. KnowledgeBest practices library4,200+ reusable AI skills in the Vault
4. SecurityGuardrails and safety rules"Never expose API keys, always sanitize HTML"
5. OrchestrationAI decision-making logicWhich model to use, when to escalate
6. ExecutionDeterministic scriptsHealth checks, deployments, reports

4b. The Supervisor

An autonomous monitoring system that watches over the entire AaaS ecosystem. It runs on a schedule — no human needed to start it.

ScheduleScanWhat It ChecksOutput
Every 4 hoursInfrastructure PulseAll domains up? Functions running? Content intact? (No SupraForge overwrites)Alert if anything fails
Daily 08:47Operations BriefNew signups, email deliveries, pipeline completions, error rates, Firestore statsDaily Report →
Daily 10:30Content GeneratorWhich blog collections are thin? Auto-generates 5 new entity draftsBlog submissions
Monday 08:00Context SyncRe-indexes all 5,865 skills and directives into Pinecone for semantic searchSync log
Monday 09:00Weekly ReviewCross-project infrastructure audit, stale indexes, deployment driftWeekly Report →
Wednesday 14:23Strategy ReviewGrowth metrics, market positioning, conversion analysis, build prioritiesStrategy Report →

This means: Every morning at 08:47, a report is generated showing what happened overnight — signups, emails sent, pipeline completions. Every Monday, a deeper audit runs. Every Wednesday, strategic analysis. All deployed as branded HTML pages at aaas-design.web.app.

4c. The Skill Vault

A repository of 4,200+ reusable AI capabilities — from Next.js patterns to TDD workflows to security scanning. When the system needs to build something new, it searches the Vault first (semantic search via Pinecone) rather than writing code from scratch. This dramatically reduces development time and errors.

5. Automated Content & Discovery

The content pipeline runs continuously, feeding the blog and driving SEO traffic:

GRAVITYCLAW (Research Agent) Discovers new AI tools, models, benchmarks, papers ↓ FACTORY (75 AutoResearch Agents) A/B tests hypotheses, generates content variants ↓ FIRESTORE SYNC (Cloud Function) Receives research → creates entity submissions ↓ AAAS.BLOG (Knowledge Index) Entity pages with SEO metadata, schema markup Each page includes a CTA → agents-as-a-service.com/activateSUBSCRIBER BRIDGE (Cloud Function) Blog subscriber? Auto-creates platform account Triggers context engineering pipeline ↓ EMAIL DRIP 7-stage activation campaign begins

This means: content creates itself → attracts users → converts them → onboards them → activates them. The entire funnel is autonomous.

6. The Email Intelligence Layer

Email is the primary interface for AaaS — both for activation and for the actual agent service.

Outbound: 7-Stage Drip Campaign

Each email is composed by AI using the user's context package. Not templates — personalized content generated by Vertex AI based on the user's industry, tech stack, and company stage. Delivered via Resend with RFC 8058 unsubscribe headers.

Inbound: Email-to-Agent Pipeline (Building)

The next major feature: users email agent-function@aaas.select → the system parses the task → routes to the right agent → executes → replies with results. No app needed — just email.

Tracking & Safety

FeatureStatus
Delivery tracking (delivered, opened, clicked)Live
Bounce handling (auto-disable after 3 bounces)Live
Complaint handling (auto-unsubscribe)Live
One-click unsubscribe (RFC 8058)Live
Email preferences pageLive
Resend webhook integrationNeeds Config

7. Technology Stack

Frontend
Next.js 14 + Tailwind CSS
React framework with static export for platform/name, server-rendered for blog. Turborepo monorepo with shared component library (@aaas/ui).
Backend
Firebase (Google Cloud)
Auth (multi-provider), Firestore (NoSQL database), Cloud Functions v2 (Node 20), Cloud Storage, Hosting (6 sites).
AI
Vertex AI (Gemini 2.0 Flash)
Context generation, industry classification, email personalization, agent suggestions. ~$0.001 per pipeline run.
Email
Resend
Transactional email delivery with React-compatible templates, webhook tracking, and RFC 8058 compliance.
Search
Pinecone (E5 Embeddings)
Semantic search across 5,865 skills. Multilingual-E5-Large model, 1024 dimensions. Powers skill discovery and context matching.
DevOps
AntiGravity ADE
AI-powered development engine with 6-layer architecture, 26 local skills, 9 operational directives, and autonomous build protocols.

8. Autonomous vs. Human-Required

The system is designed for maximum autonomy. Here's what runs itself and what still needs a human:

Fully Autonomous No Human Needed

User signup → context engineering → onboardingEnd-to-end automated
7-stage email activation dripAI-personalized, scheduled delivery
Infrastructure health monitoringEvery 4 hours, auto-alerts on failure
Daily/weekly/strategic reportsAuto-generated and deployed
Blog content generationGravityClaw → Factory → Blog pipeline
New feature developmentAI builds, tests, deploys autonomously
Subscriber → platform bridgingBlog readers auto-converted to users

Requires Human Action Manual Step

Azure AD app registrationMicrosoft auth provider for enterprise users
DNS record managementSPF/DMARC email authentication records
Stripe payment configurationSubscription billing and pricing tiers
API key rotationResend, Porkbun, and other external services
Strategic business decisionsPricing, partnerships, feature prioritization

Current open issues requiring human action: 7 items tracked here →

9. Operating Costs

The infrastructure is designed for minimal cost at current scale:

ServiceMonthly CostNotes
Firebase (Spark/Blaze plan)~$0–25Pay-as-you-go, free tier covers most usage
Cloud Run (blog)~$0–10Only charges for actual requests
Vertex AI (Gemini 2.0 Flash)~$1–5Context pipeline: ~$0.001 per user signup
Resend (email)Free / $20Free tier: 3,000 emails/month
PineconeFree / $70Free tier: 100K vectors sufficient for now
Domains (3)~$30/yearaaas.name, agents-as-a-service.com, aaas.blog
Claude Code (development)$100–200AI development agent for autonomous building
Total (current)~$150–300/monthScales with usage, not upfront

10. What's Next — The Roadmap

A full 42-item improvement plan exists at aaas-design.web.app/improvement-plan.html. The highest-impact items:

Phase 1 — Now
Inbound Email Pipeline
Users email agent-function@aaas.select → system processes → agent executes → replies with results. The core product experience.
Phase 2 — Next
Agent Execution Engine
Actually running agent tasks — Python runtime, skill loading from Vault, result formatting, quality scoring.
Phase 3 — Growth
Social & SEO Automation
LinkedIn auto-posting, UTM analytics, content calendar, referral system. Fully automated marketing engine.
Phase 4 — Scale
Billing & Marketplace
Stripe subscriptions, credit system, agent marketplace where users can discover and deploy specialist agents.

11. Key Links & Access Points

ResourceURL
Main Websiteagents-as-a-service.com
Login / Sign Upaaas.name
Knowledge Indexaaas.blog
All Deliverablesaaas-design.web.app/deliveries
Open Issues (Human Action)aaas-design.web.app/open-issues
Improvement Plan (42 items)aaas-design.web.app/improvement-plan
Daily Operations Reportaaas-design.web.app/daily-report
Strategy Reportaaas-design.web.app/strategy-report
Firebase ConsoleFirebase Console →
GitHub Repositoriesgithub.com/ibossyNr1