Ayala Agentic AI

A Model Context Protocol server implementation for the Ayala Agentic AI project. This enables AI agents to query for context across many sources.

Features

• A common Agentic AI reference architecture (or platform in the future) for Ayala Group
• Powered by open standard agent protocols, it enables LLM-powered agents to reliably locate context, invoke tools, and collaborate with other AI agents within the Ayala Group technology landscape
• It provides an "agentic layer" to Ayala Group data platforms such as Open Data, Concati, Inquiro, and GShare
• It unlocks potential synergies through a common AI technology framework (context sharing vs. data sharing)
• It will natively support widely used LLM clients such as ChatGPT, Claude, and Gemini

Design Principles

• Ease-of-Use: Support existing AI client apps (ChatGPT, Copilot, Claude) instead of building new ones
• Cost Efficiency: Allow web-based interface for non-license users and for cost-sensitive BUs
• Flexibility: Use open-standard agent protocols for integration (MCP, A2A, etc.)
• Synergy: Work with Ayala Group platforms and initiatives such as Concati, Inquiro, and Open Data
• Governance-First: Build AI governance requirements into the platform's foundational design rather than retrofitting compliance capabilities (Mosaic, MLFlow, etc.)

Architecture

The Ayala AI Assistant (MCP) serves as the central hub connecting various agentic models with context and tools providers across the Ayala Group ecosystem.

Components

Agentic Models:

• ChatGPT - OpenAI's conversational AI model
• Claude - Anthropic's AI assistant
• Databricks - MLOps and data science platform

Context and Tools Providers:

• Systems of Record - Core business systems and databases
• Ayala Open Data - Centralized data platform for the Ayala Group
• Concati / Inquiro - Data discovery and analytics platforms

The architecture enables seamless integration between AI models and Ayala's data ecosystem, providing a unified interface for AI-powered business operations.

Developing

Setup

Requires uv - see the project README for installation instructions.

make setup

Run using the MCP inspector

source .venv/bin/activate

mcp dev src/agentic/__main__.py

Run using an MCP client

Use this configuration in your MCP client (e.g., Claude Desktop, Cursor, etc.):

{
  "mcpServers": {
    "agentic": {
      "command": "<full-path-to-uv>",  // e.g. /Users/username/.local/bin/uv
      "args": [
        "--directory",
        "path/to/agentic",  // update this with an absolute path
        "run",
        "agentic"
      ]
    }
  }
}

Run tests

make test

Publishing

export UV_PUBLISH_TOKEN=...
make publish