Salesforce broadens headless offering for agentic AI access

In what could be a major pivot affecting the UX for longtime Salesforce customers -- while also making its data and features more accessible to developers and AI agents -- the company plans to significantly expand its Headless 360 offering.

Salesforce is tapping headless architecture, once thought of as an approach only for developer-intensive companies that wanted to heavily customize commercial cloud services. Headless is now being tested to build AI infrastructure, as agentic coding tools rapidly open the door for many more organizations to try it in pro-code and low-code environments.

Salesforce isn't alone in its thinking. Anthropic, Informatica (owned by Salesforce) and AWS are also early champions of headless as a vehicle for AI automation, loosely defined as "where the agent becomes the 'head' and then distributes data to many other channels at once."

Headless commerce has been around for some time, too, for digital sales. BigCommerce, Acquia, Shopify and other competitors have offered their versions of it. Commercetools was created as a headless platform from the start.

Salesforce had offered limited headless capabilities for Commerce Cloud, but with new feature releases to accommodate AI shopping agents, it has greatly expanded the platform's capabilities.

Breaking out of traditional interfaces and "going headless" is a strategy that permeates an organization's marketing, commerce and even supply chain operations, said Liz Miller, analyst at Constellation Research. It's more than just a single application or a bundle of cloud services. But many organizations will find that it's necessary to compete as agentic AI rewards the companies that can pivot.

Becoming an AI enterprise is not a rip-and-replace of previous technology initiatives, Miller said, but rather it builds on them.

Digital transformations were the gateway to building AI-ready infrastructure. Digital enterprises enabled customers to hop from channel to channel (such as from text to a phone app to a web browser) and pick up where they left off. Cloud-driven digital infrastructure allows customers to be everywhere, all at once. It also requires that data be stored in one place, and that applications be composable.

Now, AI developers need all of the above, especially composable architecture.

"If I want to be an AI organization that has all of these tools, all of these amazing workflows, all of these amazing outcomes for my people to be better at their jobs, and for my customers to have a better experience, I'd better be an AI enterprise," Miller said. "So that means all of these agents, all of this data, all the plumbing has to go to all the different places. That's headless [architecture]."

Headless 360 coming

Some of Headless 360 -- MCP servers and API tools that offer Salesforce functionalities to developers -- was previewed last April at the Salesforce TDx developer conference. In his TDx keynote, Salesforce co-founder Parker Harris mused, "Why should you ever log into Salesforce again?"

Salesforce revealed more of what it plans for Headless 360: Data 360 Server, an MCP server that avails Salesforce data to agents of many flavors, stripes and third-party vendors, including Cursor and Anthropic's Claude; Headless 360, which avails more than 200 Salesforce APIs to headless development; and numerous AI skills and integrations to third-party business apps.

Salesforce is betting that headless will become the new interface, despite all the brand capital it has invested in the interface over the decades and its customers tailoring their sales, service and marketing operations to work within the traditional Salesforce and Slack web fronts.

"People have invested in semantic models. They basically created these analytics pipelines," said Rohan Kumar, who joined Salesforce two months ago as president and chief platform officer after nearly 30 years at Microsoft.

"The good news, essentially, is that all the effort is actually adding value," he said. So when you talk about the knowledge layer, it's not some brand-new thing that you're doing for headless. All that headless is doing, essentially, is making it agent-readable."

Then there's the matter of securing agents so they don't run amok across networks and damage compliance programs, data stores or customer relations.

As more and more powerful agents emerge -- Salesforce Headless 360, for example, enables agents to dynamically discover what they can do, rather than developers specifying every API to access and every workflow path to take -- how can users govern agents without creating a Hugging Face incident of their own? CIOs and CISOs are tasked with keeping AI agents at heel, and need assurances.

"The IT side of the world is very concerned about managing agent sprawl -- like hundreds of agents, thousands or millions," Kumar said. "Then there is the Chief Information Security Officer side of the world, where specifically a lot of the agents today are running on behalf of a user, the OBO model. As these agents become more and more sophisticated, that could be dangerous."

Kumar said that Salesforce is addressing these concerns with an 18-character Agent ID system, which plugs into Salesforce and Data 360 governance policies, and MuleSoft Agent Fabric, an agent management layer that, in theory, can orchestrate Agentforce agents and those from third parties.

Data 360 MCP server, an expanded Data 360 API, the Agent Skills repository, the Slackbot MCP server and client, and Salesforce Multi-Framework are now generally available. Agent 360 Skills for Data 360 are coming later this month. The Headless 360 MCP server and Headless Experience Layer are in open beta.

Don Fluckinger is a seasoned B2B technology journalist with more than 30 years of experience specializing in enterprise IT, digital experience and content management. As a senior news writer at Informa TechTarget, he delivers award-winning analysis that helps IT and business leaders navigate complex technologies to enhance customer and employee experiences. Got a tip? Email him.