AI OS Documentation — Agents, Architecture & Deployment
Everything you need to understand, configure, and deploy the AI OS agentic operating system. AI OS provides a Virtual Corporate HQ where an Agent Fleet of 73 specialized sub-agents is organized into five documentation sections:
- Getting Started — plan selection, deployment, and API key setup
- Core Systems — the Agent Fleet, Hermes, Self-Improve, Analytics, Client Engine, and more
- Production Features — Media Production, Monetization, and the Batch Queue
- License Agreements — Community, Business, and Enterprise terms
- Infrastructure — API Reference, Deployment, Notifications, and Security
Getting Started
Core Systems
Agent Fleet
73 specialized sub-agents across 6 AI models, 4 routing tiers, and 11 departments. Roles, capabilities, escalation paths, and definitions.
Knowledge & Records
The governed company document library: a catalog over the Memory Vault, uploads, and agent artifacts, with reader-scoped access and retention on every record.
Hermes Agent
Always-on MCP background worker. Remote approvals, walkaway mode, cron jobs, mobile bug diagnosis, and self-improving skills.
Self-Improve
Grok Build plans upgrades to AI OS itself and can open a draft GitHub PR. Every apply and every PR always requires your explicit approval.
Reasoning & Automation
Process-verified steps whose actions cannot run unverified; verified-steps, reflexion and tree-search patterns for pipelines; and agents that author automations instead of executing tasks — measured 16× cheaper on a real task, reading primary sources.
Oversight & Approvals
The management layer: server-enforced approval gates with risk classes, batch approval, and the operator oversight ledger — decisions per day, median time-to-decision, pending-queue age, and remediation cost logging, on the Costs dashboard beside spend.
Analytics (AI-Signal-First)
First-party, cookieless analytics that show the AI crawler traffic Google Analytics can't see — per hosted site, with answer-engine referrals.
Client Engine
Email nurture sequences with built-in unsubscribe compliance, appointment booking on generated sites with calendar invites, and a drag-and-drop CRM pipeline kanban.
Web Studio Business Tools
Lead capture into the CRM, sales funnels wired to your Stripe checkout, dynamic pages from a dataset, and a per-site Site Manager dashboard.
Agent-Ready Sites
JSON-LD, llms.txt, Open Knowledge Format bundles, and Ed25519-signed content provenance — the machine-readable layer every generated site ships.
Skills & Pipelines
21 procedural skill files and declarative YAML pipeline chains. Skill anatomy, authoring, and pipeline orchestration.
AI Business Clone
A replica of one person's voice, expertise and limits, built on a shared company profile. Drafts for review, directs agents, never sends.
Knowledge Graph
Auto-categorizing knowledge base with semantic connections, radial visualization, and vault storage.
Design System
DESIGN.md dual-structure protocol, brand reasoning, WCAG linting, URL cloning, and cross-platform export.
Production Features
Media Production
Remotion video, Google Vids, and Blender 3D asset generation from natural language prompts.
Monetization
Product Factory, Lead Generation, Marketing Hub, and Golden Loop. End-to-end revenue pipelines.
Batch Queue
Mass content production at economy-tier cost. Queue management, priority scheduling, and output routing.
License Agreements
Three license tiers, Business and Enterprise both one-time purchases rather than subscriptions:
- Community — free, open-source, self-hosted
- Business — $1,997 one-time
- Enterprise — $4,997 one-time
Community License
Free, open-source core. 19 agents, 6 departments, self-hosted. Permissive license for personal and commercial use.
Business License
$1,997 one-time. All 73 agents, all 11 departments, self-instance theming, priority email support. Perpetual license.
Enterprise License
$4,997 one-time. Everything in Business plus SSO/SAML, advanced security, custom agent development.
Infrastructure
- API Reference — 40+ REST endpoints, authentication, rate limits, error codes
- Deployment — PM2 + Nginx with TLS, Docker, VPS provisioning
- Notifications — WebSocket alerts, Telegram Bot API, Slack webhooks
- Security — opt-in, report-only AI security suite
API Reference
40+ REST endpoints with request/response schemas. Authentication, rate limits, and error codes.
Deployment
PM2 + Nginx with TLS, Docker containers, VPS provisioning, and production hardening checklist.
Notifications
Real-time WebSocket dashboard alerts, Telegram Bot API, and Slack Incoming Webhooks.
Security
Opt-in, report-only AI security suite — platform self-assessment, the Web Studio publish gate, and the managed-client security service.
Frequently Asked Questions
What is AI OS?
AI OS is the agentic operating system documented here: a Virtual Corporate HQ where an Agent Fleet of 73 specialized sub-agents, across 6 AI models, 4 routing tiers, and 11 departments, handles work such as SEO audits, media production, and monetization pipelines.
How many agents does AI OS have, and how are they organized?
73 specialized sub-agents across 6 AI models, 4 routing tiers, and 11 departments. The Agent Fleet page documents each role, its capabilities, and its escalation paths.
What license options are available?
Community is free and open-source with 19 agents and 6 departments, self-hosted. Business is $1,997 one-time and offers all 73 agents, all 11 departments, self-instance theming, and priority email support. Enterprise is $4,997 one-time and adds SSO/SAML, advanced security, and custom agent development.
What does the Hermes agent do?
Hermes is an always-on MCP background worker that provides remote approvals, walkaway mode, cron jobs, mobile bug diagnosis, and self-improving skills.
How do I start using AI OS?
Follow the Getting Started guide: choose a plan, configure your API keys, and explore your Virtual HQ in under five minutes.
What is the Knowledge Graph?
An auto-categorizing knowledge base with semantic connections, radial visualization, and vault storage that serves as long-term memory for every agent.
What is the Knowledge & Records department?
A governed company document library: a catalog over the Memory Vault, extracted uploads, and agent-output artifacts, with owner, sensitivity, reader-scoped access, and retention on every record. Every agent on every tier reads it through a fenced, untrusted-by-default choke-point. See Knowledge & Records.