Skip to content

Rea Platform Components

This page documents the Rea-specific components that integrate with MCP.

Consulting Pods

Consulting Pods are multi-agent workflows that orchestrate multiple specialized AI agents to accomplish complex tasks. Each pod defines:

  • A set of specialized agents with specific roles
  • Workflow logic for agent collaboration
  • Input/output specifications
  • Configuration options

Pod Structure

class ConsultingPod
{
    public string $id;
    public string $name;
    public string $description;
    public array $agents;         // Specialized agents in the pod
    public array $config;         // Pod configuration
    public string $category;      // e.g., "research", "analysis", "creative"

    public function execute(array $inputs): PodResult;
    public function executeAsync(array $inputs): string; // Returns job ID
}

Example Pods

Pod Name Agents Purpose
Research Pod Researcher, Fact-Checker, Summarizer Deep research on topics
Analysis Pod Data Analyst, Visualizer, Interpreter Analyze data and generate insights
Content Pod Writer, Editor, SEO Specialist Create and optimize content
Code Review Pod Reviewer, Security Analyst, Performance Expert Comprehensive code reviews

MCP Integration

Consulting Pods are exposed via MCP as:

Tools:

  • rea_create_pod - Create a new consulting pod
  • rea_run_pod - Execute a pod with inputs
  • rea_list_pods - List available pods

Resources:

  • rea://pods - List all pods
  • rea://pods/{id} - Get pod details and recent executions

Prompts:

  • rea_pod_builder - Interactive prompt for designing new pods

Command Room

The Command Room is Rea's activity capture and analysis system. It collects user activity data from the Chrome extension and provides:

  • Activity Logging: Records user actions across web applications
  • Pattern Detection: Identifies repetitive tasks and workflows
  • Automation Suggestions: Recommends automations based on patterns
  • Context Awareness: Provides relevant context for AI agent decisions

Activity Structure

class CommandRoomActivity
{
    public string $id;
    public string $userId;
    public string $app;           // Application name (e.g., "gmail", "notion")
    public string $action;        // Action type (e.g., "click", "type", "navigate")
    public array $metadata;       // Action-specific data
    public Carbon $timestamp;
    public ?string $sessionId;    // Group related activities
}

Data Flow

┌──────────────┐    ┌──────────────────┐    ┌───────────────────┐
│   Chrome     │───►│   Command Room   │───►│   Pattern         │
│  Extension   │    │   Service        │    │   Analyzer        │
└──────────────┘    └────────┬─────────┘    └─────────┬─────────┘
                             │                        │
                             ▼                        ▼
                    ┌────────────────┐       ┌───────────────────┐
                    │   Activity     │       │   Automation      │
                    │   Database     │       │   Suggestions     │
                    └────────────────┘       └───────────────────┘

MCP Integration

Command Room is exposed via MCP as:

Tools:

  • rea_search_command_room - Search activities by query, date range, app

Resources:

  • rea://command-room/recent - Recent activity feed
  • rea://command-room/patterns - Detected patterns

Prompts:

  • rea_automation_analysis - Analyze activities and suggest automations

n8n Workflow Integration

Rea integrates with n8n for cross-application workflow automation. MCP extends this by:

  • Allowing AI agents to trigger n8n workflows
  • Exposing workflow status and results
  • Enabling workflow discovery and parameter inspection

MCP Integration

Tools:

  • rea_trigger_workflow - Trigger an n8n workflow with parameters
  • rea_get_workflow_status - Check execution status

Resources:

  • rea://workflows - List available workflows
  • rea://workflows/{id} - Workflow details and execution history

LLaMA Agent Engine

Rea uses LLaMA (Large Language Model Meta AI) as its primary language model, enhanced with:

Reflection Prompt Framework

A 5-step reasoning framework that improves accuracy:

  1. Understand - Restate the user's request
  2. Plan - Outline steps to fulfill the request
  3. Execute - Carry out the plan (including tool calls)
  4. Verify - Check the response addresses the request
  5. Reflect - Consider if there's a better approach

Tool Integration

LLaMA's tool calling is mapped to MCP:

// MCP tool definition format
$mcpTool = [
    'name' => 'notion.search',
    'description' => 'Search Notion pages',
    'inputSchema' => [
        'type' => 'object',
        'properties' => [
            'query' => ['type' => 'string']
        ],
        'required' => ['query']
    ]
];

// Converted for LLaMA
$llamaTool = [
    'type' => 'function',
    'function' => [
        'name' => 'notion.search',
        'description' => 'Search Notion pages',
        'parameters' => $mcpTool['inputSchema']
    ]
];

Chrome Extension

The Rea Chrome Extension captures user activity for the Command Room:

Captured Data

  • Page visits and navigation
  • Click actions and form interactions
  • Time spent on applications
  • Cross-app workflows

Privacy Controls

  • User-configurable capture settings
  • Domain whitelist/blacklist
  • Data encryption in transit
  • Local processing options

MCP Relevance

The extension provides context that enhances MCP tool usage:

  • Recent activity informs tool suggestions
  • Patterns help predict user intent
  • Cross-app context improves automation recommendations

Next Steps