What is ClaudeForce?#

ClaudeForce is a partnership between Salesforce and Anthropic that brings Claude — the “number one AI” — together with Salesforce, the “number one CRM.” The pitch is simple: take your trusted business data, workflows, and governance, and put them to work inside Claude, governed by the exact same permissions and business rules your company already runs on.

The first product out of the gate is Salesforce in Claude, a plugin that ships with 37 pre-built skills for sellers, grounded in 27 years of Salesforce experience. It’s available today for pilot customers, with an open beta planned for September 2026.

This isn’t a standalone product. It’s the latest chapter in a collaboration that’s been building for a while — Claude Tag in Slack, Claude in Agentforce, and Claude as a launch partner for Slack Code. ClaudeForce is the umbrella that ties all of it together under one partnership.

Salesforce in Claude: 37 skills, zero UI#

The most interesting claim on the page is a philosophical one: “a shift from software as the interface to software powering every interface.”

For decades, SaaS gave us fixed screens and menus. The information was always in there — but assembling it was your job. Open the list, click into the account, read back through the history, hold it all in your head, and only then decide what to do. A seller could lose an entire morning before making a single call.

Salesforce in Claude flips that. Claude reasons across your systems and brings the answer and the action to you, instead of making you navigate to find it. In practice, the 37 skills cover things like:

• Asking a deal question and getting grounded answers from live revenue data
• Building an account plan in seconds
• Running multi-step workflows without leaving Claude
• Keeping the pipeline updated automatically

Headless 360 and MCP: the part developers should care about#

Underneath the marketing, the architecture is what makes this genuinely novel. Salesforce is building on Headless 360 — a collection of Salesforce capabilities (data, apps, workflows, agents, and the governance around them) that AI can call directly over MCP.

For years, that value was locked behind the interface. It lived in your data, your relationships, and your rules, but the only way to reach it was to log in and click through screens. MCP is what frees it — an AI agent can work with Salesforce wherever the user already is: in Slack, in Claude, or in the native Lightning interface, with the same data and the same answers.

Here’s the detail that matters most for builders: the same capabilities are open to any developer who wants to build their own integration. ClaudeForce simply gives you a ready-made experience — a plugin you install instead of a stack you assemble by hand.

Guardrails without a new permissions model#

The classic enterprise objection to agentic AI is governance: “who is this agent, and what is it allowed to do?” ClaudeForce’s answer is that you don’t build a new permissions model at all.

Every answer and every action runs through your existing Salesforce permissions and business rules. Claude sees what the user is authorized to see, and can do what that user is authorized to do — no more. Nothing new to stand up, nothing to re-audit, nothing to configure account by account. An admin connects Salesforce in Claude once, and it works for the whole team from day one.

On the write side, the controls are just as concrete:

• Claude can check with you before emailing anyone outside your company
• When it updates a record, it changes only the field it said it would

The framing is clean: humans direct, agents execute, Salesforce governs.

Enterprise Frontier Safeguards#

Salesforce also worked with Anthropic on Enterprise Frontier Safeguards, which combine the privacy of zero data retention with state-of-the-art safeguards for detecting serious misuse. Eligible customers will be able to keep activity data in cloud infrastructure they control, under their own encryption keys, access policies, and audit logging. When monitoring flags a pattern, the alert goes directly to the customer — with no Anthropic human review required.

It rolls out in phases beginning fall 2026.

The roadmap: from Sales to the whole platform#

Sales is just the first persona. The roadmap covers essentially the entire Salesforce Cloud lineup, all marked “coming soon”:

My take#

Two things stand out to me as someone who builds on both Salesforce and agentic AI.

First, MCP as the unlock is the real story. The interface was always the bottleneck — not the data, not the model. Exposing Salesforce capabilities over MCP turns the CRM from a destination you visit into a capability other systems can call. That’s a much bigger deal than “a chatbot for sales.”

Second, governance as a feature, not an afterthought. The single strongest line on the page is that Claude “sees what the user is authorized to see, and can do what that user is authorized to do, no more.” For enterprise adoption, that’s the difference between a pilot and a production rollout.

The “AI CRO for every seller” framing is marketing, but the architecture underneath is real. If Salesforce in Claude delivers on the MCP promise, we’re looking at the first credible path to agents that can actually act on the system of record — not just chat about it.

I’ll be watching the September 2026 open beta closely.