Quick start guide: Connect your AI agent to Pearl’s MCP Server
- Pearl Team
- Jun 6, 2025
- 1 min read
Updated: Jun 24, 2025
Supercharge your AI agent with real-time access to verified human Experts using the Model Context Protocol (MCP). This guide walks you through the fastest way to get up and running.
Step 1: Prerequisites
Node.js 18+ installed
An MCP-compatible AI agent (e.g. Claude Desktop, Cursor IDE, VS Code, or custom agent using OpenAI Agents SDK)
An API key from Pearl (request one at:api@pearl.com)
Step 2: Connect to the remote MCP Server
You’ll use the mcp-remote adapter to bridge your local client to Pearl’s remote MCP Server via Server-Sent Events (SSE).
Install
npm install -g @modelcontextprotocol/mcp-remote
Start the adapter
mcp-remote --url https://mcp.pearl.com/mcp --header "Authorization: Bearer YOUR_API_KEY"
Connecting from MCP client (e.g. Claude Desktop)
Step 3: Discover available tools
Once connected, your agent will automatically discover the following tools:
Tool | Description |
askPearlAi | Fast AI-only response from Pearl’s proprietary LLM |
askPearlExpert | AI intake → escalates to a real human Expert |
askExpert | Immediate handoff to a verified Expert |
getConversationStatus | Query ongoing interactions |
getConversationHistory | Retrieve session history |
These tools follow standard MCP discovery and JSON schema conventions.
Step 4: Handle authentication
Pearl’s MCP Server utilizes Bearer token authentication for secure communication. You don’t need to manage tokens manually - just include your API key in the Authorization header when initializing the connection.
If using an SDK, make sure your transport layer supports passing custom headers.
Step 5: Invoke a tool
Here's a minimal example using the OpenAI Agents SDK (Python):
Step 6: Test in a live environment
Once integrated:
Trigger expert escalation based on model confidence or domain rules
Route legal/medical/technical queries to human Experts
Comments