The Fable 5.1 Architecture Standard

Stop Letting AI Blindly Guess Code.
Orchestrate with Invariant Gates.

Traditional coding agents rush to write lines of code without considering tacit assumptions. FableFlow puts Claude Fable 5.1 as the Master Architect: decomposing requirements into strict contracts, delegating to worker subagents, and gating delivery behind adversarial invariant checks.

Architect Input & Presets

Interactive

Select a high-friction engineering problem or type your custom requirement:

Task Requirement Prompt

Simulation Mode:

Architect: Claude Fable 5.1 (Planning & Spec)

Worker Lane: GPT-5.6 / Sonnet (Implementation)

QC Gate: Clean-Context Fable 5.1 (Invariant Audit)

Live 4-Stage Execution Visualizer

Ready for Execution

STAGE 1

Architect Decomposition & Tacit Invariants

Owner: Claude Fable 5.1

Click 'Run FableFlow Architect Pipeline' to generate contract...

STAGE 2

Worker Subagent Task Allocation

Owner: Implementation Lane

Awaiting Stage 1 contract sign-off...

STAGE 3

Adversarial Invariant Verification Gate

Owner: Fable QC Gate

Awaiting worker subagent patch...

STAGE 4

Certified Delivery & Audit Worklog

Owner: Provenance Engine

Awaiting Stage 3 verification pass...

1-CLICK INTEGRATION

Connect FableFlow to Your Coding Agent Mid-Task

Zero-dependency Model Context Protocol (MCP) server. Fully compatible with Claude Code, Cursor, Codex, and Antigravity.

claude mcp add fable-flow -- npx -y github:jastfan/fable-flow

// In .cursor/mcp.json:
{
  "mcpServers": {
    "fable-flow": {
      "command": "npx",
      "args": ["-y", "github:jastfan/fable-flow"]
    }
  }
}

// In mcp_config.json:
{
  "mcpServers": {
    "fable-flow": {
      "command": "npx",
      "args": ["-y", "github:jastfan/fable-flow"]
    }
  }
}

git clone https://github.com/jastfan/fable-flow.git
cd fable-flow
node mcp-server.js

DEEP ARCHITECTURAL COMPARISON

Why Invariant-Gated Architecture Wins

Compare raw unconstrained agents vs complex paid orchestrators vs FableFlow 5.1.

Feature / Capability	Raw Coding Agent (No Contract)	Traditional Orchestrators	⚡ FableFlow 5.1
Inference & Billing Model	⚠️ Uncontrolled token consumption	❌ Proprietary SaaS fee + markup	✔ BYOK: Direct Anthropic/OpenAI/Ollama (0% Markup)
Tacit Invariant Detection	❌ Blind (Only follows happy path)	⚠️ CLI prompt instructions only	✔ Adversarial Extraction & Enforcement
Interactive Visual Studio	❌ None (Terminal / Chat only)	❌ None (Raw Python CLI)	✔ High-Contrast 4-Stage Live Studio
Adversarial Bug Injection Testing	❌ Hallucinates tests verifying flawed code	❌ Static checks only	✔ Real-time Invariant Gate Failure Detection
Tooling Protocols Supported	⚠️ Proprietary chat UI	⚠️ Custom Python script runners	✔ Universal MCP (Claude, Cursor, Antigravity)
Automated CI/CD PR Gates	❌ None	❌ None	✔ 1-Click GitHub Actions Invariant Gate

INTEGRATION & EXPORT HUB

Export Invariants to Your Tool of Choice

Take the verified invariants and rules and drop them straight into your team's workflow.

📋 Cursor Rules (.cursorrules)

Export Fable 5.1 Architect mode instructions as a system prompt for Cursor IDE.

🤖 Claude Code Skill (SKILL.md)

Production Claude Code skill definition for autonomous multi-agent task execution.

🚀 GitHub Actions Gate (.yml)

Zero-cost CI/CD pipeline step that blocks pull requests if tacit invariants fail.

🛡️ Provenance Audit JSON

Download the machine-readable verification worklog of the current simulation.