Salesforce shipped its core CRM MCP servers to general availability earlier in 2026, but the Data 360 MCP server — released as a developer preview in May 2026 — is a different and arguably more interesting beast. Instead of pointing an agent at sObjects and Flows, it opens the Data 360 layer (formerly Data Cloud): the unified, harmonised customer data that sits across your systems. Here's what that distinction means and how to use it.
CRM rows vs. unified data
The standard Salesforce MCP server is great for transactional work — read an Opportunity, update a Case, run a Flow. Data 360 is built for a different question: "what does everything we know about this customer say?" It harmonises data from many sources into unified profiles, calculated insights, segments and data graphs, often through zero-copy connections to warehouses like Snowflake and BigQuery. The Data 360 MCP server exposes that semantic layer to an agent, so you can ask about a customer's full cross-system picture in plain English, rather than stitching together a dozen object queries yourself.
Why this is a big deal for AI agents
An agent is only as good as the data behind it, and most CRM agents are quietly limited by living inside a single object model. Data 360 changes the ceiling: the agent reasons over a harmonised, deduplicated view of the customer instead of one app's slice. That's the difference between "list this account's open cases" and "which high-value segments are showing early churn signals this month, and why." It's also why the preview matters now — teams already invested in Data Cloud get an AI-native query surface over data they've spent years unifying.
Setting up the preview
As a developer preview, expect the surface to evolve. The path mirrors the GA hosted servers: enable the Data 360 MCP server in Setup, create an External Client App for OAuth 2.0 with PKCE (mandatory), grant the "Access MCP servers" and refresh-token scopes, and add it to your client as a remote OAuth connector. Your Data 360 permissions and data-space access apply automatically, so the agent inherits exactly the visibility the connecting user has — never more. Being a preview, keep it out of anything load-bearing until it reaches GA. See MCP config file location.
Verify and secure
Ask: "Summarise the unified profile for this account and any calculated insights on it." A coherent cross-source answer confirms you're hitting the Data 360 layer rather than raw CRM. On security: unified customer data is among the most sensitive an enterprise holds, so connect with a least-privilege user, scope data-space access tightly, and start read-only before granting anything that writes segments or activations. See MCP security best practices and memory privacy for AI agents.
Going further
Data 360 sits naturally alongside warehouse servers — see Snowflake MCP setup and BigQuery MCP setup, and the Snowflake agent profile. For analyst workflows, wire it into a data-analyst loadout or browse the data-analysis category. New to Salesforce on MCP? Start with the core Salesforce MCP setup.