Salesforce has today expanded Headless 360 with an MCP-based architecture that lets AI agents discover and act on existing capabilities across enterprise workflows.
The move takes headless CRM beyond making Salesforce functionality available outside its interface, giving agents access to customer data, business logic, and governed actions at runtime.
CX leaders must now consider whether their customer data and governance frameworks can support reliable agent-led execution.
“Salesforce was built on the idea that customers should be able to extend the platform far beyond the interface we provide," noted a Salesforce executive.
"Headless 360 carries that idea into the agentic era. Instead of rebuilding data, logic, and governance for every new agent, customers can activate what they have already built on Salesforce wherever work happens.”
The Limits of a Connected CRM
The original Headless proposition was built around separating CRM capabilities from the Salesforce interface, allowing data and functionality to be accessed across different environments.
Today, it has expanded to address one of the pressing limitations that becomes more evident as AI agents take on more customer-facing work.
For example, an agent may have access to Salesforce but still reaches the boundary of what has been explicitly connected or configured for it, causing it to stall and hand the task back to an employee.
Furthermore, accessing CRM capabilities is only useful if agents can work with the customer context required to make appropriate decisions and actions.
With data work often involving sourcing fields, mapping schemas, building segments, and activating information across tools and teams, processes that can take create severe delays.
For customers, these unnecessary gaps in Headless 360 can result in slower resolutions, inconsistent experiences, and interactions that lack the context needed to feel relevant or trustworthy.
Inside Salesforce’s Agent-Ready Headless Architecture
Most notably, Salesforce’s Headless 360 MCP Server is now available in open beta, improving on the earlier model by allowing agents to discover the Salesforce capability they need at runtime.
Rather than requiring developers to predefine every API or feature, agents can now discover, understand, and invoke Salesforce capabilities dynamically.
As a metadata-aware server, the MCP enables agents to inherit existing object relationships, validation rules, and automations alongside the user’s Salesforce own permissions and policies, allowing them to work within already established business logics.
Secondly, Salesforce is extending the architecture across its wider CX portfolio, giving agents access to capabilities spanning customer engagement, sales and data integration.
In service, Headless capabilities can bring dispatching, Field Service scheduling, and support workflows into external applications and other customer touchpoints.
For marketing, the Marketing Engagement MCP allows marketers to manage automation, customer journeys, and data extensions through natural language inside Slack.
