Pearl MCP Server

Created By

Pearl-com8 months ago

A Model Context Protocol (MCP) server implementation that exposes Pearl's AI and Expert services through a standardized interface

# mcp

# human-in-the-loop

Overview Content Tools Comments

Content

Pearl MCP Server

A Model Context Protocol (MCP) server implementation that exposes Pearl's AI and Expert services through a standardized interface. This server allows MCP clients like Claude Desktop, Cursor, and other MCP-compatible applications to interact with Pearl's advanced AI assistants and human experts.

Features

Support for both stdio and SSE transports
Integration with Pearl API for AI and expert assistance
Session management for continuous conversations
Multiple interaction modes:
- AI-only mode for quick automated responses
- AI-Expert mode for AI-assisted human expert support
- Expert mode for direct human expert assistance
Conversation history tracking
Stateful session management

Prerequisites

Python 3.12 or higher
Pearl API Key (Contact Pearl to obtain your API key)
pip or uv package manager

Installation

Clone the repository:

git clone https://github.com/Pearl-com/pearl_mcp_server.git
cd pearl_mcp_server

Create a virtual environment and activate it:

python -m venv .venv
source .venv/bin/activate  # On Windows: .venv\Scripts\activate

Install dependencies:

pip install -e .

Configuration

Create a .env file in the src directory:

PEARL_API_KEY=your-api-key-here

Running the Server

Local Development

Start the server using either stdio (default) or SSE transport:

# Using stdio transport (default)
pearl-mcp-server --api-key your-api-key

# Using SSE transport on custom port
pearl-mcp-server --api-key your-api-key --transport sse --port 8000

Using Remote Server

Pearl provides a hosted MCP server at:

https://mcp.pearl.com/mcp

This can be used directly with any MCP client without installing the Python application locally.

Available Tools

The server provides the following tools:

ask_pearl_ai
- Quick AI-only responses without human review
- Best for general inquiries and non-critical situations
- Parameters:
  - question: The user's query
  - chat_history (optional): Previous conversation context
  - session_id (optional): For continuing conversations
ask_pearl_expert
- AI-assisted human expert support
- Best for complex topics requiring expert verification
- Parameters: Same as ask_pearl_ai
ask_expert
- Direct human expert assistance
- Best for complex or sensitive topics
- Parameters: Same as ask_pearl_ai
get_conversation_status
- Check the status of an active conversation
- Parameter: session_id
get_conversation_history
- Retrieve full conversation history
- Parameter: session_id

Expert Categories

Pearl's MCP server provides access to a wide range of expert categories. The appropriate expert category is automatically determined by Pearl's API based on the context of your query, ensuring you're connected with the most relevant expert for your needs.

Here are the main categories of expertise available:

Medical & Healthcare
- General Medicine
- Dental Health
- Mental Health
- Nutrition & Diet
- Fitness & Exercise
- Veterinary Medicine
Legal & Financial
- Legal Advice
- Tax Consultation
- Financial Planning
- Business Law
- Employment Law
- Real Estate Law
Technical & Professional
- Software Development
- IT Support
- Computer Repair
- Electronics
- Mechanical Engineering
- Home Improvement
Education & Career
- Academic Tutoring
- Career Counseling
- Resume Writing
- Test Preparation
- College Admissions
- Professional Development
Lifestyle & Personal
- Relationship Advice
- Parenting
- Pet Care
- Personal Styling
- Interior Design
- Travel Planning

Each expert category can be accessed through the ask_expert or ask_pearl_expert tools. You don't need to specify the category - simply describe your question or problem, and Pearl's AI will automatically route your request to the most appropriate expert type based on the context.

Connecting with MCP Clients

Local Connection (stdio transport)

For connecting to a local MCP server using stdio transport, add the following configuration to your MCP client:

{
    "pearl-mcp-server": {
        "type": "stdio",
        "command": "pearl-mcp-server",
        "args": ["--api-key", "your-api-key"],
        "env": {
            "PEARL_API_KEY": "Your Pearl Api Key"
        }
    }
}

Remote Connection using mcp-remote

Some MCP clients don't support direct connection to remote MCP servers. For these clients, you can use the mcp-remote package as a bridge:

Prerequisites:
- Node.js 18 or higher
- npm (Node Package Manager)
Configuration for remote server:

{
    "mcpServers": {
        "pearl-remote": {
            "command": "npx",
            "args": [
                "mcp-remote",
                "https://mcp.pearl.com/sse"
            ]
        }
    }
}

Configuration file locations:
- Claude Desktop:
  - Windows: %APPDATA%\Claude\claude_desktop_config.json
  - macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
- Cursor: ~/.cursor/mcp.json
- Windsurf: ~/.codeium/windsurf/mcp_config.json
Additional Options:
- Force latest version: Add @latest to npx command
```
"args": ["mcp-remote@latest", "https://mcp.pearl.com/sse"]
```
Troubleshooting:
- Clear stored credentials: rm -rf ~/.mcp-auth
- View logs:
  - Windows (PowerShell): Get-Content "$env:APPDATA\Claude\Logs\mcp.log" -Wait -Tail 20
  - macOS/Linux: tail -n 20 -F ~/Library/Logs/Claude/mcp*.log
- Test connection: npx mcp-remote-client https://mcp.pearl.com/sse

Custom Python Client

import asyncio
from mcp.client.session import ClientSession
from mcp.client.stdio import StdioServerParameters, stdio_client

async def main():
    # For stdio transport
    async with stdio_client(
        StdioServerParameters(command="pearl-mcp-server", args=["--api-key", "your-api-key"])
    ) as (read, write):
        async with ClientSession(read, write) as session:
            await session.initialize()
            
            # List available tools
            tools = await session.list_tools()
            print(tools)
            
            # Call Pearl AI
            result = await session.call_tool(
                "ask_pearl_ai", 
                {
                    "question": "What is MCP?",
                    "session_id": "optional-session-id"
                }
            )
            print(result)

asyncio.run(main())