I'm a bit confused between using Hosted MCP Servers and just using Data Cloud for grounding my AI agents. Both seem to provide "context" to the LLM. Is MCP intended to replace the Data Cloud integration, or do they serve different purposes when it comes to real-time 1:1 personalization in Marketing Cloud?
3 answers
They are complementary rather than competing. Data Cloud is your "Lakehouse" where you harmonize massive amounts of historical data from disparate sources (web, mobile, legacy) to create a Unified Profile. MCP, on the other hand, is a protocol for action and live connectivity. Think of Data Cloud as the "Memory" of what the customer did in the past, and the MCP Server as the "Hands" that allow the AI to reach into a system—like your CRM or a live logistics database—right now to perform a task. For real-time personalization, you'd use Data Cloud to segment the user and MCP to let the agent actually update their current case status.
So, if I wanted to create a "Next Best Action" agent, would I use Data Cloud to calculate the score and then use an MCP tool to actually present the offer to the customer? I'm trying to figure out where the hand-off happens in a standard workflow.
MCP is much lighter. If you don't need the massive scale of a CDP, a simple Hosted MCP server can give your AI enough context to be useful without the overhead of Data Cloud.
Well said, Elizabeth. For small to mid-sized orgs, starting with Hosted MCP is a much faster way to get "intelligent" responses without the complex implementation cycle of a full Data Cloud rollout.
Precisely, Joseph. Data Cloud does the heavy lifting of calculating the "Affinity Score" or "Churn Risk" based on billions of data points. Then, when the customer interacts with a chatbot, the AI agent uses an MCP server to "fetch" that score and then "executes" an action—like generating a discount code—via another MCP tool. MCP makes the integration modular so you aren't hard-coding those actions into your AI prompts.