Salesforce expands Headless Data 360 for MCP so developers can bring insights to agents
SiliconANGLE Kyt Dotson
Salesforce just opened up its Data 360 tool so AI agents can plug into customer data directly, no special interface needed. That means agents like Claude or ChatGPT can now query and build with Salesforce data just by describing what they need in plain language.
Based on reporting by SiliconANGLE, Kyt Dotson — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Salesforce dropped a new piece of its agentic AI puzzle today: Headless Data 360 for Model Context Protocol, a way to hand AI agents direct access to governed customer data without forcing them through a dedicated user interface first. It's a bigger deal than the name suggests, because it changes who — or what — gets to touch a company's data infrastructure.
The mechanics matter here. Data 360 exposes more than 200 existing APIs as programmable endpoints, meaning agents can build, transform, map, segment and activate data fields on their own. Previously, anyone working with that data had to already know the exact fields, schemas and tables involved before they could do anything useful — a process Salesforce says could stretch into days or weeks as people bounced between tools and teams just to pull together what they needed.
This builds on Headless 360, which Salesforce launched back in April as what it called the "front door" into its ecosystem, letting developers wire up conversational interfaces across the platform via MCP. Data 360 pushes further into the data layer itself. A business user can now type something like a request for lifetime value across electronics purchases while excluding software, and the agent handles the rest — writing the query, building the calculation, constructing the semantic models and relationships behind it. No exact field names required, no precise function calls. Just the idea and the context, handled through whichever agent someone prefers, whether that's Claude, Cursor, ChatGPT or Salesforce's own Agentforce.
Dr. Paramjit "Romi" Chopra, who runs the Midwest Institute for Minimally Invasive Therapies, offered a real-world data point: his agents can now reach unified patient context through MCP without needing a custom-built interface for every single use case. He also flagged something worth paying attention to — the underlying metadata is what he called a "physician-curated ontology," meaning humans structured the responses even when a generative AI system is the one answering. That's the trust layer Salesforce is leaning on.
The Data 360 MCP Server itself is Salesforce-hosted and comes bundled with prepackaged Skills for modeling, mapping, transforms and activation, meant to eliminate the grunt work of building these capabilities from scratch. Custom Skills will let users encapsulate their own repeated workflows too. Those Skills arrive later in August, so the full picture of how this plays out in production is still a few weeks away.
My take — AI-written commentary, not fact-checked reporting
Salesforce keeps building the plumbing before most companies even know they need it, and that's the smart play — whoever owns the data access layer for agents ends up owning the leverage, regardless of which chatbot brand wins the popularity contest. The Chopra example is the one to watch, not the marketing copy about hours instead of months; healthcare data governed by a human-curated ontology is exactly the kind of guardrail that determines whether agentic AI becomes trustworthy infrastructure or just another hype cycle with a permissions problem.
Read more about this at: SiliconANGLE