Salesforce Agentic Identity Will Register Outside AI Agents

Salesforce plans to require every outside AI agent that touches its platform to register first, under a system called Salesforce Agentic Identity. According to Crypto Briefing, the requirement covers external agents that connect through Salesforce APIs or through its Model Context Protocol servers, and it is scheduled to launch in November 2026.

The move is one of the clearest signs yet that the enterprise software world is shifting from open agent access to governed agent access. Agents will no longer be able to blend in with ordinary integrations. Each one will carry its own identity, its own permissions, and a meter running on its activity.

How Salesforce Agentic Identity will work

Under the new rules, each registered agent receives its own unique identity and its own set of permissions, according to Crypto Briefing. Those permissions will be narrower than what a human user typically receives, a deliberate constraint meant to limit what an autonomous system can reach inside a customer org.

Customers will also be able to monitor what their registered agents access, giving IT teams a record of agent activity that many organizations currently lack. The rollout is part of a wider set of upgrades Salesforce is introducing under its Headless Toolkit and AIforce banners, which cover software that reaches into a platform directly, without going through the usual on-screen interface.

The tooling arrives slightly ahead of the deadline. According to Crypto Briefing, Salesforce's Winter '27 release, available from October 12, 2026, will include API Catalog features that customers can use to register their agents.

Deadlines depend on how agents connect

The rollout comes in stages, and the deadlines depend on who the customer is and how the agent connects. New customers who buy Salesforce services after September 17, 2026, must register their external agents within three months of being notified that Salesforce Agentic Identity is available.

Existing customers get more runway for API-based agents: they will need to comply when their contracts come up for renewal. Anyone using Salesforce's MCP servers, whether a new or existing customer, must register within three months of receiving notice.

Traditional integrations that are not agent-driven are untouched. According to analysts tracking the change, pricing and security options for conventional integrations remain unchanged, which draws a sharp line between human-built software and autonomous agents.

The meter behind the registration desk

Registration is only half the story. Salesforce plans to link the system to a new billing model that charges for agent activity through Flex Credits, according to reports on the AIforce rollout. Each successful call a registered agent makes over MCP or a direct API will be metered as a Headless Platform Interaction.

The price is not set yet. The Headless Platform Interaction multiplier is listed as to be announced, and Salesforce has said it will give 30 days' notice before metering begins. Organizations must also be on Flex Credit billing to participate, which means some customers will need to change how they buy before they can register at all.

For budgeting teams, that uncertainty is the practical problem. The registration requirement is a compliance task, but the meter turns agent traffic into a recurring cost line, and the rate card is still blank. Analysts covering the rollout advise organizations with renewals in the next six months to complete an agent inventory now, while the multiplier is still unpublished and nothing is being charged.

Why agent identity is becoming the enterprise battleground

Salesforce is not moving alone. Microsoft has built Guardian Agents to supervise enterprise AI agents as insiders, a parallel effort to bring autonomous systems under the same kind of oversight once reserved for employees. The broader pattern is unmistakable: vendors that once competed on how easily agents could connect are now competing on how tightly agents can be governed.

The stakes are concrete. An unregistered agent with broad credentials can read customer records, trigger workflows, and move data between systems, all without a human clicking anything. A registered agent with narrow permissions and a monitored activity log is a far smaller risk, and a far easier one to switch off.

That trade-off explains the staged deadlines. Salesforce cannot simply flip a switch on thousands of existing integrations without breaking customer workflows, so it is using renewals and notice periods to ease organizations into the new model. The MCP deadline is tighter because MCP servers are purpose-built agent on-ramps, and Salesforce wants those governed first.

For agent builders, the message is equally direct. Building an agent that talks to Salesforce will soon mean designing for registration, narrow permissions, and metered calls from the start. Agents that assume open, unmetered access will need rework, and vendors that cannot say which agents they run, who owns them, and what they touch will find that conversation harder at renewal time.

Sources: Crypto Briefing, Asha Pura Softech on Salesforce AIforce, Vantage Point on agentic Salesforce traffic. Related coverage: Microsoft Guardian Agents Treat Enterprise AI Agents as Insiders, Nvidia Nemotron Bolt Puts Enterprise Agents on Local Machines.