MCP Platform MCP Server

MCP Platform

A scalable MCP-based execution platform that allows users to interact via a chat UI, with an LLM (via LlamaIndex) interpreting intent and controlled execution of application-specific actions via MCP.

Architecture Overview

┌─────────────┐
│   Chat UI   │  React-based frontend
└──────┬──────┘
       │
       ▼
┌─────────────────────────────┐
│  Orchestrator (AI Gateway)  │  FastAPI - Conversation management, LLM orchestration
└──────┬──────────────────────┘
       │
       ├────────────────────┐
       ▼                    ▼
┌─────────────┐      ┌─────────────┐
│     LLM     │      │ MCP Client  │
│ (LlamaIndex)│      │             │
└─────────────┘      └──────┬──────┘
                            │
                            ▼
                     ┌─────────────┐
                     │ MCP Server  │  Tool registry, auth, audit, routing
                     └──────┬──────┘
                            │
          ┌─────────────────┼─────────────────┐
          ▼                 ▼                 ▼
    ┌───────────┐    ┌───────────┐    ┌───────────┐
    │    HR     │    │    ERP    │    │  DevOps   │
    │  Domain   │    │  Domain   │    │  Domain   │
    └───────────┘    └───────────┘    └───────────┘

Key Principles

• Separation of Concerns: Each component has a single responsibility
• Domain Isolation: Applications are isolated with no cross-domain calls
• Configuration-Driven: New domains added via configuration, not code changes
• LLM is Advisory, MCP is Authoritative: LLM suggests actions, MCP enforces them

Components

Frontend (Chat UI)

• Captures user input
• Displays assistant responses
• No business logic
• No direct LLM or MCP access

Orchestrator (AI Gateway)

• Manages conversation state
• Interfaces with LLM via LlamaIndex
• Supplies tool definitions to LLM
• Parses structured tool calls
• Invokes MCP Client

MCP Server

• Registers and exposes MCP tools
• Enforces authorization
• Routes calls to application domains
• Audits all executions
• No LLM or UI logic

Application Domains

Each domain contains:

• Tool definitions (namespaced, e.g., hr.get_employee)
• Adapter implementation
• Permission model
• Configuration

Included domains:

• HR: Employee lookup, department info
• ERP: Invoices, inventory management
• DevOps: Kubernetes operations, logs, scaling

Quick Start

Prerequisites

• Python 3.10+
• Node.js 18+
• Docker (optional)

Local Development

1. Clone and setup:

cd mcp_poc
cp .env.example .env
# Edit .env with your LLM API keys

2. Install Python dependencies:

pip install -e ".[dev]"

3. Start MCP Server (terminal 1):

export PYTHONPATH=$PWD/src
python -m mcp_server.main

4. Start Orchestrator (terminal 2):

export PYTHONPATH=$PWD/src
python -m orchestrator.main

5. Start Frontend (terminal 3):

cd frontend
npm install
npm run dev

6. Open browser: http://localhost:3000

Using Docker

# Build and run all services
docker-compose up --build

# Access:
# - Frontend: http://localhost:3000
# - Orchestrator API: http://localhost:8000
# - MCP Server API: http://localhost:8001

Testing with Mock LLM

For testing without LLM API keys, use the mock provider:

# config/settings.yaml
llm:
  provider: mock

API Endpoints

Orchestrator (port 8000)

MCP Server (port 8001)

Tool Examples

HR Domain

# Get employee information
hr.get_employee(employee_id="E001")

# Search employees
hr.search_employees(department="Engineering", query="developer")

# List departments
hr.list_departments()

ERP Domain

# Get invoice
erp.get_invoice(invoice_id="INV-001")

# Create invoice
erp.create_invoice(
    customer="Acme Corp",
    items=[{"description": "Service", "quantity": 1, "unit_price": 1000}]
)

# Check low stock
erp.check_low_stock(category="Components")

DevOps Domain

# List pods
devops.list_pods(namespace="production")

# Get pod logs
devops.get_pod_logs(pod_name="api-server-xyz")

# Scale deployment
devops.scale_deployment(deployment_name="api-server", replicas=3)

Adding a New Domain

1. Create domain directory:

src/domains/mydomain/
├── __init__.py
└── config.yaml (optional)

2. Implement adapter