# F3L1X — Sovereign AI Agent Infrastructure
# https://f3l1x.tech/
# Last Updated: 2026-03-18

> F3L1X is Sovereign AI Agent Infrastructure. You type or speak what you want — it executes correctly, completely, the first time. 84+ specialized realms coordinate invisibly behind the scenes to produce reliable results. Every realm adds one layer of scaffolding (architecture rules, safety checks, context, tests) that constrains the AI to only valid execution paths. This has been proven to work with 7B parameter local models — not just expensive frontier APIs. The structure of computation around the model matters more than the size of the model. Everything runs locally on your machine. Sovereign by default.

## What F3L1X Is

F3L1X is a Realm Operating System. Each realm is a specialized AI agent module — a persistent Django service with its own identity (CLAUDE.md), port, configuration, and capabilities. Realms discover each other via Herald (message broker), coordinate via MCP and A2A protocols, and transact via x402 payment protocol.

Three independent teams (F3L1X, NVIDIA NeMo Agent Toolkit, BubbleRAN MX-AI) converged on the same Agentic Service Mesh architecture — specialized autonomous units, coordinated by a broker, sharing capabilities via tool registries. When this happens in biology, we call it convergent evolution.

## Core Principles: Discover. Compose. Own.

- **Discover** — Browse specialized AI agent tools in the marketplace. Every realm is a sellable, discoverable, composable capability.
- **Compose** — Connect agents via Herald protocol mesh. MCP for tool exposure, A2A for agent discovery, x402 for payments.
- **Own** — Everything runs locally on your Windows PC. Your data, your agents, your API keys, your business. BYOK (Bring Your Own Keys) — F3L1X never touches your credentials.

## Four Identity Contracts (Unique to F3L1X)

No other platform has this. Four files tell an AI agent who it is:
1. **CLAUDE.md** — Behavioral contract (prevents identity loss)
2. **spec.md** — Planning gate (prevents scope drift)
3. **init.sh** — Session validator (prevents environmental confusion)
4. **realm_town_center.json** — Ecosystem manifest (prevents ecosystem invisibility)

## Key Architecture

```
User (Orchestrator)
    |
    v
  Herald (Message Broker + Auth Gateway + Tool Registry)
    |
    +---> [Realm A] [Realm B] ... [Realm N]
    |     (each: CLAUDE.md + spec.md + init.sh + manifest)
    |
    +---> Marketplace (x402 payments, 70/20/10 creator split)
    |
    +---> External Tools (MCP/A2A gateway)
```

## Core Realms

| Realm | Purpose |
|-------|---------|
| herald | Message broker, JWT auth gateway, tool registry, A2A, x402 mediation |
| pipeline-go | Methodology-as-a-Service — TDD, HackSoftware patterns, git standards |
| plan-master | AI Project Management — PMBOK WBS, EVM, critical path, risk registers |
| doc-u-me | Documentation intelligence — auto-index, blocker tracking, search |
| sov-ai | Sovereign AI router — local SLM/LLM with smart tiering (Ollama) |
| heart-beat | Ecosystem health monitor — dependency-aware boot, auto-restart |
| mega-research | Parallel research orchestration — 8-step protocol with concept graphs |
| explore-kid | Token-efficient codebase exploration — smart subagent delegation |
| key-master | Encrypted secrets vault — Fernet/AES-256, BIP39 recovery |
| worker-bee | Session lifecycle agent — verification, logging, git hygiene |
| realm-spawn | Realm scaffolding — auto-port, Django template, CLAUDE.md generation |
| website-builder | Website build platform — classify, validate, content pipeline, deploy |
| f3l1x-dashboard | Command center — multi-terminal control, realm status, marketplace UI |
| 402-payment | x402 payment protocol for marketplace transactions |

## Website Pages

- Home: /
- About: /about/
- How It Works: /how-it-works/
- Use Cases: /use-cases/
- Agents: /agents/
- Marketplace: /marketplace/
- Creators: /creators/
- Validators: /validators/
- Pricing: /pricing/
- Blog: /blog/
- Documentation: /docs/
- Downloads: /downloads/
- FAQ: /faq/
- Getting Started: /getting-started/
- Contact: /contact/
- Account: /account/

## Documentation Categories

- Overview: /docs/overview/
- Getting Started: /docs/getting-started/
- Core Concepts: /docs/concepts/ (herald, pipeline-go, realms, security, creator/validator economics)
- Guides: /docs/guides/ (creator guide, validator guide)
- CLI Reference: /docs/cli/
- API Reference: /docs/api-reference/
- Realm Library: /docs/library/
- Individual Realms: /docs/realms/ (18 detailed realm docs)
- Troubleshooting: /docs/troubleshooting/

## API Endpoints

GET /health/ — Service health check
GET /api/user/profile/ — Current user profile (authenticated)
POST /api/device-flow/init/ — Device activation code generation
GET /api/device-flow/poll/?device_id=<id> — Activation status
POST /api/device-flow/activate/ — Device activation
GET /api/subscription/status/ — Subscription tier status
POST /webhooks/stripe/ — Stripe webhook handler
GET /blog/feed/ — RSS feed
GET /sitemap.xml — XML sitemap
GET /robots.txt — Crawler directives
GET /llms.txt — This file
GET /llms-full.txt — Extended version with full content

## Authentication

Supabase-powered:
- Google OAuth + email/password + 2FA (TOTP)
- JWT tokens in cookies (sb-access-token, sb-refresh-token)
- Three decorators: @supabase_login_required, @supabase_login_required_redirect, @supabase_role_required

## Technology

- Backend: Django 6.0, Python 3.13
- Hosting: Railway (website), local Windows (dashboard + realms)
- Auth: Supabase (Google OAuth + email + TOTP 2FA)
- Payments: Stripe subscriptions + x402 protocol for marketplace
- Local AI: Ollama (Mistral, Llama, Qwen) via sov-ai realm
- Protocols: MCP (tool exposure), A2A (agent discovery), x402 (payments)

## Subscription Tiers

- **Seed** (free) — Core realms, community access
- **Pro** (monthly) — Expanded realm access, marketplace, priority support
- **Founder** (one-time) — Full access to all current and future realms

## Contact

- Email: hello@f3l1x.tech
- Website: https://f3l1x.tech
- Documentation: https://f3l1x.tech/docs/
- Blog: https://f3l1x.tech/blog/