API Reference
Data CatalogGet a DemoStart Free
Start typing to search…

OpenAI Implementation Guide

Learn how to integrate AgentSource MCP with OpenAI's API to build AI agents that can access real-time business and contact data.

Overview

This guide demonstrates how to use AgentSource MCP (Model Context Protocol) with OpenAI's API. MCP enables your AI agents to seamlessly access Explorium's data through a standardized protocol, allowing for dynamic tool discovery and execution.

Prerequisites

  • Python 3.8 or higher
  • OpenAI API key
  • AgentSource API key
  • Basic understanding of OpenAI's API

Installation

First, install the required package:

pip install openai

Quick Start

1. Initialize the OpenAI Client

from openai import OpenAI

# Initialize the client with your OpenAI API key
client = OpenAI(api_key="your-openai-api-key")

2. Configure AgentSource MCP

Create a response with MCP tools configured:

response = client.responses.create(
    model="gpt-4.1",
    tools=[{
        "type": "mcp",
        "require_approval": "never",
        "server_label": "explorium",
        "server_url": "https://mcp.explorium.ai/sse",
        "headers": {
            "Authorization": "Bearer YOUR_AGENTSOURCE_API_KEY"
        }
    }],
    input="Your query here"
)

3. Basic Example

Here's a complete example that searches for companies:

from openai import OpenAI

# Initialize client
client = OpenAI(api_key="your-openai-api-key")

# Make a request to find banks
response = client.responses.create(
    model="gpt-4.1",
    tools=[{
        "type": "mcp",
        "require_approval": "never",
        "server_label": "explorium",
        "server_url": "https://mcp.explorium.ai/sse",
        "headers": {
            "Authorization": "Bearer YOUR_AGENTSOURCE_API_KEY"
        }
    }],
    input="Find 10 banks in the US with less than 5000 employees that use Azure"
)

# The response will contain the results
print(response)

Common Use Cases

Finding Companies with Specific Criteria

# Search for tech companies using specific technologies
response = client.responses.create(
    model="gpt-4.1",
    tools=[{
        "type": "mcp",
        "require_approval": "never",
        "server_label": "explorium",
        "server_url": "https://mcp.explorium.ai/sse",
        "headers": {
            "Authorization": "Bearer YOUR_AGENTSOURCE_API_KEY"
        }
    }],
    input="Find software companies in California with 100-500 employees using AWS"
)

Finding Contacts at Specific Companies

# Get engineering directors at Microsoft
response = client.responses.create(
    model="gpt-4.1",
    tools=[{
        "type": "mcp",
        "require_approval": "never",
        "server_label": "explorium",
        "server_url": "https://mcp.explorium.ai/sse",
        "headers": {
            "Authorization": "Bearer YOUR_AGENTSOURCE_API_KEY"
        }
    }],
    input="Get 50 directors of engineering at Microsoft along with their email addresses"
)

Analyzing Specific Departments

# Analyze VPs in a specific department
response = client.responses.create(
    model="gpt-4.1",
    tools=[{
        "type": "mcp",
        "require_approval": "never",
        "server_label": "explorium",
        "server_url": "https://mcp.explorium.ai/sse",
        "headers": {
            "Authorization": "Bearer YOUR_AGENTSOURCE_API_KEY"
        }
    }],
    input="Inspect all VPs in the Engineering department at Salesforce and highlight those working on Agentforce"
)

Understanding the Response

The MCP integration automatically handles:

  • Tool discovery and listing
  • Parameter formatting
  • API calls to AgentSource
  • Response parsing

Your agent will receive structured data that it can process and present in a user-friendly format.

Advanced Configuration

Multi-turn Conversations

For follow-up queries in the same context:

# Initial query
response1 = client.responses.create(
    model="gpt-4.1",
    tools=[{
        "type": "mcp",
        "require_approval": "never",
        "server_label": "explorium",
        "server_url": "https://mcp.explorium.ai/sse",
        "headers": {
            "Authorization": "Bearer YOUR_AGENTSOURCE_API_KEY"
        }
    }],
    input="Find all VPs at Tesla"
)

# Follow-up query
response2 = client.responses.create(
    model="gpt-4.1",
    tools=[{
        "type": "mcp",
        "require_approval": "never",
        "server_label": "explorium",
        "server_url": "https://mcp.explorium.ai/sse",
        "headers": {
            "Authorization": "Bearer YOUR_AGENTSOURCE_API_KEY"
        }
    }],
    input="Now get their email addresses",
    previous_response_id=response1.id
)

Tool Approval Settings

You can control whether tools require approval before execution:

  • "never" - Tools execute automatically
  • "always" - Tools require user approval
  • "on_error" - Approval required only on errors

Debugging and Monitoring

Response Tracing

To understand the execution flow, you can inspect the response object which contains:

  • Tool calls made
  • Input parameters sent
  • Responses received
  • Agent's interpretation

Example Response Structure

# Access the response outputs
for output in response.output:
    if output.type == "mcp_call":
        print(f"Tool called: {output.name}")
        print(f"Arguments: {output.arguments}")
        print(f"Result: {output.output}")
    elif output.type == "message":
        print(f"Agent response: {output.content}")

Best Practices

  1. API Key Security: Never hardcode API keys in your code. Use environment variables:

    import os

client = OpenAI(api_key=os.getenv("OPENAI_API_KEY"))

  2. Clear Queries: Be specific in your queries for better results:

    • ✅ "Find 20 CTOs at fintech companies in New York with 50-200 employees"
    • ❌ "Find some tech people"

  3. Handle Rate Limits: Implement appropriate error handling and retries

  4. Optimize Requests: Batch related queries when possible to reduce API calls

Available Tools

Through MCP, your agent automatically has access to all AgentSource tools including:

  • Company Search: Find businesses based on various criteria
  • Contact Discovery: Locate professionals with specific roles
  • Email Enrichment: Get contact information
  • Company Matching: Match and enrich company data
  • Prospect Matching: Match and enrich prospect data
  • And more: The full suite of AgentSource capabilities

The beauty of MCP is that tools are discovered dynamically - you don't need to explicitly define them.

Error Handling

try:
    response = client.responses.create(
        model="gpt-4.1",
        tools=[{
            "type": "mcp",
            "require_approval": "never",
            "server_label": "explorium",
            "server_url": "https://mcp.explorium.ai/sse",
            "headers": {
                "Authorization": "Bearer YOUR_AGENTSOURCE_API_KEY"
            }
        }],
        input="Your query"
    )
except Exception as e:
    print(f"Error: {e}")

Next Steps

  • Explore the complete AgentSource MCP documentation
  • Check out implementations using other frameworks (LangGraph, Python SDK)
  • Learn about advanced filtering and search capabilities
  • Understand rate limits and quotas

Support

If you encounter any issues or have questions:

Note: Remember to replaceYOUR_AGENTSOURCE_API_KEY and your-openai-api-key with your actual API keys.