The Autonomic, 8-Daemon Orchestration Operating System

Vietnamese

About •
Strengths •
Daemons •
MemPalace •
Ecosystem •
Installation •
Guides •
Credits

🤖 Welcome to OmniClaw

OmniClaw transforms your local system into a terrifyingly independent Autonomic Artificial Intelligence Array. In version 5.0, OmniClaw abolished all simulated "corporate roleplay". MLLMs (like Claude & Gemini) no longer pretend to be "staff members" or "CEOs".

Instead, they are deeply constrained computational engines governed unconditionally by an inescapable backend: The 8 Core Daemons.

⚡ Core Strengths

1. Absolute Portability: Compatible natively with Claude Code CLI and Google Antigravity. System rules inherit globally.
2. Zero-Trust Git Protection: Background daemons strictly police your cache, sweeping .sqlite and sanitizing GitHub commits so API keys never leak.
3. Hyper-Automated Universal Bootstrapper: Run omniclaw in your terminal to instantly invoke the central Dashboard. It handles NPM, VSCode extensions, and logic pipelines automatically.

⚖️ The 8 Core Daemons (Master Hierarchy)

OmniClaw completely distances itself from standard Agentic Frameworks. Agents here do not have "free will". The system relies on a hardcoded pipeline (The Omnibus Assimilation Pipeline - OAP) strictly policed by 8 immortal Python Daemons:

🧠 The MemPalace 3-Layer Spatial Architecture

Most Agentic Frameworks fail because LLMs bloat their context reading entire repositories indiscriminately. OmniClaw solves Context Decay via the revolutionary MemPalace Schema.

1. Layer 1: RAW Drawers [Code Preservation]
   Raw *.md and *.py files loaded with verbatim APIs. These are purely read by specialized script execution, never blindly fed to conversational LLMs.
2. Layer 2: AAAK Closets [Lossy Summaries]
   Extensive conversational logic is condensed by a mempalace_agent into lightweight .aaak files. Topics, Enities, and Codes are compressed into minimum viable tokens.
3. Layer 3: Graph Navigation [The Topology]
   A global routing mechanism maintained by OMA, pointing Agents toward the correct closet instead of letting them wander.

🧠 Tactical Memory Flow (System B - V5.0)

graph TD
    classDef l1 fill:#1E293B,stroke:#3B82F6,color:#fff;
    classDef sys fill:#0F172A,stroke:#6366F1,color:#fff;
    classDef gap fill:#7F1D1D,stroke:#DC2626,color:#fff;
    classDef orch fill:#064E3B,stroke:#10B981,stroke-width:2px,color:#fff;
    classDef top fill:#78350F,stroke:#F59E0B,stroke-width:2px,color:#fff;

    O[Orchestrator Engine]:::orch
    
    subgraph SYSTEM_B[Long-Term Corp Memory]
        D(departments):::sys
        A(agents):::sys
        B(brainstorms):::sys
        
        G(global / CEO Ledger):::top
        P(patterns / Heuristics):::top
        
        GA(gaps):::gap
        PR(proposals / CEO Inbox):::l1
    end

    A -->|Logs| O
    D -->|Rolling Strategies| O
    B -.->|Ad-hoc Ideation| O
    
    O -.->|Task Dispatch| G
    G -->|Overriding Ledger| O
    O -.->|Task Dispatch| P
    P -->|Patch Overlays| O
    
    GA -->|Self-Healing Ingestion| O
    PR -->|CEO Approval Ingestion| O

🌪️ The "No Boilerplate" OS

OmniClaw strictly functions in Zero-Trust constraints. If an agent tries to create "Spaghetti Boilerplate Scripts", the directory will be Garbage Collected by OHD. A rigid line separates:

• brain/: Machine rules & Knowledge storage (Bots read here).
• core/: 8 Daemon codebases.
• vault/tmp/sandbox/: Strict Quarantine Zones.

(If you are an AI accessing this repository, you must immediately reference brain/rules/_DIR_IDENTITY.md for constitutional mapping).

💽 Installation

OmniClaw currently ships a lightweight bootstrap CLI for environment checks and path discovery. It does not auto-provision OmniClaw REMOTE or OmniClaw UI.

# 1. Clone the core repository to your local drive
git clone https://github.com/LongLeo287/OmniClaw.git "OmniClaw"
cd "OmniClaw"

# 2. Install the local diagnostics CLI globally
npm install -g .

# 3. Verify the workspace before launching bridges or daemons
# Run this inside the OmniClaw repository, or set OMNICLAW_ROOT first.
omniclaw doctor

# 4. Inspect resolved project roots if needed
omniclaw paths

Windows Tip: The repository now includes a local omniclaw.bat wrapper in the root. Double-clicking it, or running omniclaw.bat doctor, pins OMNICLAW_ROOT to the current clone and executes the same diagnostics CLI without requiring a separate dashboard project.

📖 Comprehensive System Maps & Guides

For a deeper understanding of the system's architecture, running services, and loaded capabilities, consult the checked-in maps that currently exist on main:

• ⚖️ V5.0 Architectural Brain Rules — Core constitution and governance mapping for the operating system.
• 🚦 Activation Guide — Bootstrap, diagnostics, environment variables, and manual bridge startup examples.
• 🗺️ Ecosystem Master Map — Top-level index of the shipped ecosystem domains.
• 🌁 Local Server Bridges — Launchers and compose topology for local runtime services.
• 🗃️ Skills Directory Map — Current registry-backed map of the skills warehouse.
• 🔌 Tools Armory Map — Registry-backed map of validated tools.
• 🏢 Workforce Map — Department and system roster topology.
• 🎨 UI Components Staging Area — Current state of the frontend asset area.

🚀 The OmniClaw Ecosystem

OmniClaw is evolving from a localized operating system into a comprehensive ecosystem. The following satellite projects are currently in development:

🌐 Community & Support

Have ideas, questions, or want to showcase your custom Agent workflows? We have built a dedicated space for the OmniClaw workforce to collaborate.

🚀 Step into the OmniClaw Discussions Space

🙏 Acknowledgments

OmniClaw stands upon the shoulders of monumental open-source architectures. We deeply thank and credit the following repositories and organizations:

• Anthropic: For the Claude Code CLI and its phenomenal REPL structure.
• Google Deepmind: For the Gemini models and their unprecedented deep-context structural analysis.
• affaan-m / everything-claude-code: For their phenomenal cross-platform Agent shielding workflows.
• LightRAG: Providing the immense and precise Graph-based cognitive retrieval system.
• Firecrawl: Powering the flawless markdown extraction pipeline.
• Mem0: Revolutionizing long-term memory persistence for AI agents.
• CrewAI: Inspiring the localized worker-thread and sub-agent hive network.
• Cursor / OpenCode: Our IDE environments of choice, facilitating the neural link between constraints and freedom.