Skip to main content

​
What is MCP?

The Model Context Protocol (MCP) is an open standard for connecting AI agents to external data and tools. Developed by Anthropic and adopted across the industry, MCP is becoming the common language enterprises use to expose data to AI systems—regardless of which vendor’s agents are consuming it. Think of MCP as a universal adapter. Instead of building custom integrations for every AI platform, you build one MCP server and any compliant agent can use it. This matters for enterprises because:
  • Standardization — Define once how your data is accessed, enforce it everywhere. Your CDP, product catalog, and internal knowledge base speak the same protocol whether Gradial, your internal tools, or other vendor agents are calling them.
  • Interoperability — As more platforms adopt MCP, your agents can communicate with each other. A Gradial agent executing a content workflow can hand off to another vendor’s agent for specialized tasks, passing context seamlessly.
  • Future-proofing — Investing in MCP means your data infrastructure works with tomorrow’s AI tools, not just today’s. The protocol is designed to evolve with the ecosystem.
  • Control — You decide exactly what data and capabilities to expose. MCP servers sit in your infrastructure, behind your security controls, serving only what you explicitly permit.
Gradial’s MCP integration lets you connect any MCP-compliant server—whether you’ve built it yourself, deployed an open-source implementation, or are using a vendor’s MCP endpoint—and immediately extend what Gradial agents can do.

​
Why Connect an MCP Agent?

MCP integrations let you extend what Gradial agents can do without waiting for native integrations. Common use cases include:
  • First-party data access — Connect your CDP, data warehouse, or customer database so agents can personalize content using real audience data
  • Internal tool access — Expose internal APIs, proprietary calculators, or business logic to agents
  • Custom content sources — Pull from internal knowledge bases, product catalogs, or pricing systems
  • Workflow automation — Trigger actions in systems that don’t have native Gradial integrations

​
Prerequisites

Before you begin, make sure you have:
  1. A running MCP server that implements the MCP specification
  2. The server’s base URL and endpoint path
  3. Authentication credentials (if required)
  4. Admin or Integrations permissions in Gradial
Your MCP server must be accessible from Gradial’s infrastructure. If your server is behind a firewall, you’ll need to allowlist Gradial’s IP addresses. Contact support for the current list.

​
Using Your MCP Agent

Once connected, your MCP agent becomes available to Gradial agents automatically. The agent capabilities description helps Gradial decide when to use your MCP tools. To explicitly invoke your MCP agent in a prompt or workflow: 
Use the {Integration Name} agent to [describe the task].
Example: 
Use the HubSpot MCP Agent to pull the top 5 customer segments 
by engagement score for the Q4 campaign.

​
Example: First-Party Data MCP

A common pattern is connecting a first-party data MCP to personalize content at scale. Setup:
  • Integration Name: Customer Data Platform
  • Base URL: https://cdp-api.yourcompany.com
  • Endpoint Path: /mcp
  • Capabilities: “Query customer segments, retrieve audience attributes, look up product affinity scores, and fetch personalization rules by segment ID.”
Use case: When creating landing page variants, Gradial agents can query your CDP to understand which segments exist, what attributes define them, and which messaging resonates with each—then generate on-brand content tailored to each audience.