Salesforce’s original Headless 360 pitch was that AI agents shouldn't need dashboards built for people. That claim is worth revisiting now that the underlying architecture is starting to take shape. Salesforce has announced an expansion of Headless 360 with new Model Context Protocol (MCP) servers, Data 360 capabilities and headless experiences across Data, Sales, Service, Marketing, Commerce, MuleSoft and Informatica, letting authorised agents discover and act across the enterprise without a bespoke integration for every use case.
From applications to a governed capability layer
An enterprise application traditionally bundles an interface, data and business logic together. Headless 360 separates those layers, so an agent can reach the underlying capability directly, without going through the interface, while still inheriting the permissions and governance already built into the business.
The release spans a Headless 360 MCP Server, a Data 360 MCP Server, more than 100 reusable Agent Skills, and expanded functionality across each of those clouds. Salesforce says the MCP Server is metadata-aware, allowing agents running in Agentforce, Claude, ChatGPT and Cursor to discover what is available to them, along with the workflows, permissions and validation rules that already define how the business operates, rather than requiring an integration team to recreate that logic for every new agent.
Why Data 360 matters for CX
An agent cannot deliver a genuinely context-aware interaction if the relevant customer information stays locked inside separate applications. According to Salesforce, the new Data 360 MCP Server expands trusted customer context beyond the Salesforce interface, exposing close to 200 API operations that agents can use to access and act on customer data through natural-language interactions. Salesforce had already applied a similar headless principle to data management through Informatica.
Control has to travel with the data
Once agents can act on data and workflows from outside the traditional interface, that interface can no longer be the main point of control, so governance has to travel with the data instead. As agents move from pilots into production, the harder problem becomes deploying many of them reliably, not proving what one can do. Reusable, governed capabilities are Salesforce's answer to that. Engine, a customer service company, says it built an AI support agent in 12 days using the platform. It is only one customer example, rather than evidence of a broader pattern, but it offers an early indication of what this approach could mean for deployment speed.
CX platforms are becoming capability layers
The bigger question extends beyond Salesforce. Every CRM and customer data vendor now faces the same test: can their capabilities be discovered and used by agents across multiple interfaces, or does their value risk being hidden behind someone else's orchestration layer? This is a shift already visible in analytics, where conversational interfaces are beginning to replace the dashboard.
This expansion is the first real evidence of what the headless enterprise looks like as infrastructure. Salesforce is effectively trying to turn the platform from a destination that people and agents visit into capabilities that agents can discover and invoke wherever work happens. This is no longer just a theory about what enterprise software might look like in an agentic future. Salesforce is beginning to build the infrastructure that could make the headless enterprise a reality.