AGI Cognitive Server

Model Context Protocol Python Version

An enterprise-ready Model Context Protocol (MCP) server providing programmatic access to causal discovery, persistent homology ($B_1$ topological calculations), epistemic Gaussian Process updates, Landauer thermodynamic limits, and local HCHL (Hypergeneralized Causal-Homological Latent) inference.

Researchers and agents can mount this server directly to run complex quantitative and topological simulations locally or in the cloud.

---

🛠️ Features

• Causal Discovery: Programmatic PC and FCI causal structure recovery algorithms.
• Topological Data Analysis (TDA): 1-skeleton persistent homology cycle detection and filtration.
• Epistemic GP Belief Updating: Gaussian Process regression tracking for model beliefs and uncertainty prediction.
• Thermodynamic Auditing: Automated Landauer heat dissipation limit and computational complexity risk indicators.
• Quantized LoRA Tuning: Meta-learning matrices simulation with INT8 scaling factors.
• Wasserstein DRO & MDL: Robust optimization simulations under uncertainty radius $\epsilon$ and Kolmogorov AST code model complexity.
• HCHL Core Inference: Direct stdio pipeline calling the hypergeneralized local transformer.
• ISO/IEC & NIST Auditing: Real-time regulatory standard scoring compliance outputs.

---

🚀 Setup & Installation

1. Requirements

Ensure you have Python 3.12+ (or 3.14+) installed. Clone the repository and install requirements:

pip install -r requirements.txt

2. Sibling Dependency Note

This server acts as a gateway interface. It automatically detects and binds to parent/sibling submodules inside the main compute-intelligence-orchestrator project structure (e.g. agi-cognitive-agent-core, automated-artificial-general-intelligence, post-exotic-research-compendium).

If running standalone, ensure these sibling directories are available in your path or set PYTHONPATH:

export PYTHONPATH="/path/to/compute-intelligence-orchestrator/agi-cognitive-agent-core:/path/to/compute-intelligence-orchestrator/automated-artificial-general-intelligence"

---

🔌 Connection Setup

Add the following configuration blocks to connect this server to your preferred LLM host client.

Claude Desktop Integration

Modify your claude_desktop_config.json (typically located in %APPDATA%/Claude/claude_desktop_config.json on Windows or ~/Library/Application Support/Claude/claude_desktop_config.json on macOS):

{
  "mcpServers": {
    "agi-cognitive-mcp-server": {
      "command": "python",
      "args": [
        "c:/Users/svillalobosgonzalez1/Documents/GitHub/compute-intelligence-orchestrator/agi-cognitive-mcp-server/mcp_server.py"
      ],
      "env": {
        "APCA_API_KEY_ID": "your_alpaca_key_id",
        "APCA_API_SECRET_KEY": "your_alpaca_secret_key"
      }
    }
  }
}

Cursor Integration

1. Go to Cursor Settings -> Features -> MCP.
2. Click + Add New MCP Server.
3. Fill in details:
   • Name: agi-cognitive-mcp-server
   • Type: stdio
   • Command: python c:/Users/svillalobosgonzalez1/Documents/GitHub/compute-intelligence-orchestrator/agi-cognitive-mcp-server/mcp_server.py

---

🧪 Verification

You can verify that the server is working and resolving correctly by running the test suite:

pytest test_cognitive_mcp.py