Speak the Language of Headless 360
Learning Objectives
After completing this unit, you’ll be able to:
- Define the key terms and concepts behind Salesforce Headless 360.
- Explain how APIs, MCP, and ADLC Skills work to make Salesforce accessible anywhere.
- Identify the role of the Salesforce Metadata Catalog in grounding AI agents.
The Glossary of Headless 360 Terms
Becoming adept at a new technology is a lot easier when you have the right information. Before you start mapping out your strategy, you need to understand the concepts. This unit gives you explanations of key Salesforce Headless 360 terms so you can navigate it confidently and clearly.
Headless means decoupling the Salesforce backend intelligence, its data, logic, and APIs from any fixed UI or screen. Traditionally, Salesforce was accessed through a browser-based interface (like Lightning Experience). Going headless means that a powerful backend can now be used from anywhere, a custom app, a chat tool, a developer IDE, or an AI agent without requiring the standard Salesforce interface. Simple version: Salesforce capabilities are no longer tied to a single front door. They’re accessible from anywhere.
The Headless Experience Layer (HXL) separates your business logic from the screen it appears on. Instead of building a separate app for every surface, you define your intent once inside Salesforce and HXL renders it natively wherever your users already work—Slack, WhatsApp, a custom portal, or anywhere else. One definition. Any surface.
APIs (application programming interfaces) are the connectors that let external apps, agents, and tools talk to Salesforce—requesting data, triggering workflows, or taking action without anyone opening the Salesforce UI. If you’re already familiar with APIs, the key shift with Headless 360 is that any app or agent can now call Salesforce directly, not just Salesforce-native tools.
MCP (Model Context Protocol) is the open standard that lets any AI agent call Salesforce tools. Think of it as a universal plug instead of each agent needing a custom, one-off integration with Salesforce, MCP provides a standardized way for any agent to connect, call tools, query data, and take action.
The Salesforce Metadata Catalog is a structured index of everything inside your Salesforce org—every object, field, relationship, workflow, and permission—expressed in a way AI agents can read and understand. Instead of guessing at your data model, an agent consults the catalog to know what’s there, what it’s allowed to touch, and what the rules are before taking any action. Think of it as the org’s table of contents for AI.
Platform-hosted MCP servers are Salesforce-managed servers that expose your Salesforce APIs, data, and logic tools any AI agent can call. What makes this powerful is that there’s no integration code to write, no servers to stand up, and all Salesforce trust controls are automatically enforced on every call. Salesforce runs the infrastructure, and trust is enforced automatically. You don’t need to build or maintain the connection yourself—this is the difference between a months-long integration project and being up and running in days.
The Salesforce CLI (command line interface) is a developer tool that lets you interact with Salesforce directly from a terminal or command line, creating, deploying, and managing Salesforce components without using a browser. With Headless 360, CLI capabilities are expanding to support agentic development workflows.
An IDE (integrated development environment) is a software application where developers write, test, and deploy coding tools like Agentforce Vibes IDE, VS Code, Cursor, or Windsurf. With Headless 360, your developers no longer need to leave their preferred IDE to build Salesforce-powered experiences. They can connect to Salesforce through MCP Servers and Agent Development Lifecycle (ADLC) Skills directly from the tools they already use.
ADLC Skills are open-source skills that enable coding agents, like Claude Code or Cursor, to build, test, and observe Agentforce agents directly from any IDE. ADLC Skills are the Salesforce answer to a simple question: How do developers who live in Claude Code or Cursor efficiently build Agentforce agents without ever switching tools? ADLC has three core skills.
- Develop: Build Agentforce agents from any IDE.
- Test: Validate agent behavior without leaving your tools.
- Observe: Monitor and optimize agent performance.
There’s no Salesforce login required. Developers get enterprise safety and Salesforce frameworks built in without specialized platform expertise.
Deterministic versus probabilistic: Building a successful agent strategy means knowing the difference between deterministic and probabilistic logic. Here’s how these two concepts shape the way your AI operates behind the scenes.
- Deterministic: The agent behaves like a traditional workflow, following a defined, predictable path where the same input always produces the same output. (Think of this like a calculator: 2+2 always = 4.)
- Probabilistic: The agent uses AI reasoning to decide what to do. It can handle ambiguity and variation but outputs can differ based on context. (Think of this like a weather report: Saying it will rain really means there’s a 95% chance of rain; 5% chance of different weather.)
Salesforce Headless 360 supports both. ADLC Skills help developers build agents that combine the reliability of deterministic logic with the flexibility of probabilistic AI.
Skills represent the specific tasks an agent is trained to perform. Examples of these tasks include looking up orders, processing refunds, sending emails, and pulling reports. This functionality can be compared to a multi-tool pocket knife, where each individual tool—such as a knife, scissors, or screwdriver—is designed to perform one specific function well. Similarly, when an agent is equipped with various skills, it can select the appropriate one to successfully complete a given job.
A2A stands for agent to agent, which describes the process of one agent handing off work directly to another agent. The handoff occurs automatically without the need for a human intermediary. This interaction creates a single, seamless workflow where multiple agents can complete tasks without waiting.
Orchestration is how multiple agents and systems are coordinated to complete a complex task. Think of it like a conductor leading an orchestra. Each agent (instrument) plays its part and orchestration ensures they work together in the right order, at the right time. In Headless 360, Salesforce provides built-in cross-product orchestration so agents can work across Agentforce Service, Agentforce Sales, Data 360, and external systems seamlessly, without custom glue code holding it together.
The Salesforce Layers
Salesforce gives you already integrated, already running systems inside your enterprise. Developers are not starting from scratch. They are building on infrastructure that is already running inside their enterprise, with the data, the workflows, the permissions, and the engagement layer already in place. That is what nobody else can offer.
Layer | Product | What It Does |
|---|---|---|
System of context | All your trusted business data, unified, real-time, and ready for agents to act on | |
System of work | Decades of business logic and workflows across sales, service, and every other function, now orchestrated by agents | |
System of agency | Build, deploy, and manage agents at scale across every channel | |
System of engagement | Where humans and agents come together to get work done |