Salesforce Rebuilds Platform for AI Agents

Salesforce announced "Headless 360," its most significant architectural transformation in 27 years, designed for AI agents. This initiative exposes every platform capability via APIs, MCP tools, or CLI commands, enabling AI to manage the entire system without a browser.

Unveiled at its TDX developer conference in San Francisco, the launch includes over 100 new tools and skills immediately available to developers. Salesforce stated its decision two and a half years prior was to rebuild the platform specifically for agents, making it fully programmable and accessible from anywhere.

The "Headless 360" initiative is built on three pillars: building any way you want, deploying on any surface, and building trustworthy agents at scale. This includes over 60 new MCP tools and 30 pre-configured coding skills for external agents, native React support, and the Agentforce Experience Layer for cross-platform deployment.

Furthermore, new lifecycle management tools for testing, evaluation, and orchestration are introduced, alongside the open-sourced Agent Script domain-specific language. This aims to address the challenge of agent brittleness and ensure deterministic outcomes in enterprise AI deployments.

Salesforce is also embracing openness by integrating with major AI models like OpenAI, Anthropic, and Google Gemini, while hedging its bets on communication protocols by supporting APIs, CLIs, and MCP. This strategic shift is accompanied by a move to consumption-based pricing for Agentforce.

This transformation repositions Salesforce, moving from a traditional UI-centric model to an agent-first architecture. The company is betting that its decades of accumulated enterprise data and workflows, now accessible programmatically, will remain valuable even as AI agents gain more autonomy.