Salesforce Headless 360 exposes every major Salesforce capability as an API, MCP tool, or CLI command, enabling both coding agents and business agents to interact with the platform. The post explains two categories of MCP tools: those for coding agents (like Claude Code or Cursor) that can read metadata, generate Apex, and deploy code, and those for business agents that serve end users from surfaces like Slack, ChatGPT, or Teams. Key changes include multi-framework UI support (React alongside LWC), a Headless Experience Layer for rendering capabilities across any surface, and governance enforcement (sharing rules, field-level security, validation rules) that applies regardless of entry point. The author argues that existing Salesforce platform expertise — data models, permission architecture, business logic — becomes more valuable in an agentic world because agents inherit this accumulated context rather than requiring it to be rebuilt.
Table of contents
The platform provides context and capabilityTwo kinds of MCP toolsHow software is built, delivered, and used is changing all at onceWhat Headless 360 actually isHeadless 360 in practiceWhy most agentic prototypes don’t reach productionWhat comes nextRelated ReadingAbout the AuthorSort: