Salesforce Launches Headless 360 To Enable Agentic Workflows

Salesforce announced Headless 360, a platform-level shift that makes "everything on Salesforce" callable by agents as an API, MCP tool, or CLI command. The company framed the change as a move from UI-first workflows to agent-first operations, giving AI agents live access to Customer 360 data, workflows, and governance. The launch bundles updated tooling including Agentforce Vibes 2.0, an experience layer, and safety and production controls, and highlights Claude Sonnet 4.5 and GPT-5 among supported models.

Salesforce positions Headless 360 as three coordinated innovations. The platform exposes capabilities programmatically so agents can invoke business logic without human UI interaction.

• •New MCP tools and coding skills, providing coding agents direct, live access to organization metadata and business processes. Salesforce cites 60 new MCP tools and 30 preconfigured coding skills as immediate building blocks.
• •A new experience layer, intended to render rich native interactions across Slack, voice channels, WhatsApp, custom apps, and other surfaces, decoupling the UI from the platform.
• •Agent governance and runtime controls, enabling prelaunch testing, staged rollout, and behavior constraints for agents operating in production.

Agentforce Vibes is now available as a Vibe-coding IDE with multi-model support and prebuilt integrations for the Salesforce CLI, organization metadata, and CI/CD pipelines via a DevOps Center MCP. The default LLM in the Vibes environment is Claude Sonnet 4.5, but multi-model support explicitly includes GPT-5 and other agent frameworks. The platform emphasizes programmatic DevOps, natural language devops pipelines, and the ability for agents to run commands and mutate org state using the same metadata and APIs human developers use.

This is a strategic repositioning. Instead of competing solely on UI features, Salesforce is betting that the next wave of enterprise automation will come from agentic systems that require direct access to trustworthy system-of-record context. By exposing its CRM, service, marketing, and commerce capabilities as programmable primitives, Salesforce aims to become the control layer for enterprise agents. That creates a big advantage because most enterprises already have critical customer context inside Salesforce; agents that can call that context directly will avoid the contextual gaps that plague many third-party agent integrations.

However, the change surfaces two nontrivial problems. First, security, governance, and audit trails become more important because agents will be capable of transactional changes. Second, organizational process friction remains the primary adoption barrier. As analysts noted, exposing APIs is necessary but not sufficient; companies must adapt roles and controls to take advantage of agentic workflows.

Adoption metrics in large Salesforce orgs, integrations with major LLM providers and agent frameworks, and emerging best practices for agent governance and observability. Also watch for third-party tooling that standardizes MCP patterns and runtime safety for production agent behavior.