Salesforce plans to make external AI agents register before they get in

Salesforce wants every outside AI agent that touches its platform to stop at the front desk and sign in first. The company plans to require customers to register external AI agents that use its APIs or Model Context Protocol (MCP) servers. It is calling the system Agentic Identity, and customers will also be able to monitor what those agents access. Pricing for all this is still unknown, which is the part customers care about most. How Agentic Identity would work The requirement is scheduled to launch in November 2026. It is part of a wider set of upgrades Salesforce is rolling out under its Headless Toolkit and AIforce banners. “Headless” describes software that reaches into a platform directly, without going through the usual on-screen interface. An AI agent pulling customer records through an API is a good example. No human is clicking anything. MCP works like a universal adapter. It gives AI agents a standard way to connect to outside tools and data, and Salesforce runs its own MCP servers for that purpose. Under the new rules, each registered agent gets its own unique identity and its own set of permissions. Those permissions will be narrower than what a human user typically r...