On 15 September 2026, Anthropic released Salesforce in Claude in beta. The plugin, built with Salesforce, brings a seller's accounts, opportunities, and pipeline into Claude under that seller's existing Salesforce permissions. It includes 37 skills for the work account executives do daily, including account research, call prep, pipeline review, and CRM updates.
Sellers often spend hours of their day, Anthropic writes, assembling information scattered across Salesforce, email, call recordings, and Slack. With the plugin, Claude does that administrative work and updates Salesforce once the seller approves. Two connectors ship with it, Salesforce and Slack. The first time a seller opens the plugin, a setup skill builds a Claude Artifact for their role and book of business.
Each morning Claude delivers a brief with the day's meetings, deals closing soon, at-risk opportunities, and unread threads. From the brief a seller can tell Claude to push a close date, change a stage, or add a follow-up task, and Claude makes the update in Salesforce. Call prep pulls open opportunities and where each stands, what the account team discussed this week, unanswered threads, and questions still open from previous calls. If it finds stakeholders in those threads who are not in Salesforce yet, Claude adds them as contacts. Pointed at an opportunity, Claude scores the deal against the team's methodology. Asked to keep going, it drafts the business case and a dated mutual close plan, fills in the qualification fields, and adds the missing stakeholders as contact roles. All this is saved to the opportunity once it has been approved. After a call, Claude turns a transcript or the seller's notes into a follow-up email, a summary for the deal channel in Slack, and drafted opportunity updates for the seller to review. A seller can ask for a pipeline view and Claude builds an interactive dashboard. From the dashboard, a seller can tell Claude to move a close date or change a stage, and Claude updates the record in Salesforce. Sales leaders can run the same views across the team.
Anthropic writes that Salesforce remains as the system of record. Sellers sign in with their Salesforce credentials. Claude reads only what those permissions allow. By default, Claude asks the seller to approve each proposed change before it is written. On Team and Enterprise plans, Anthropic says it does not train its models on customer data by default. GitLab, Siemens, and Legora have deployed the plugin. Anthropic says 7,000 Salesforce sellers use it in their work.
Salesforce's Claudeforce page describes the same product. For decades, they write, software gave people screens, and assembling the information was the user's job. They describe a shift from software as the interface to software powering every interface. Headless 360, they write, is data, apps, workflows, agents, and the governance around them, callable over MCP, with no user interface required and no custom integration work. Claude routes every action through Salesforce so the company's rules stay in force. Their line is: Humans direct. Agents execute. Salesforce governs. On the write side, they say you decide how much autonomy Claude has. It can check before it emails anyone outside the company, or send once you are comfortable. When it updates a record, it changes only the field it said it would.
Editorial
The product they announced is teaching sellers not to open Lightning. The next thing you will want gone is the approve click. Anthropic writes that click as the default. The control Salesforce names as yours to loosen is outbound mail.
What remains is a record: named owners, fields that refuse a bad write, and a log someone else can read. Any database will do if it can refuse. A chat log cannot. Rows with no owners cannot. Headless 360 keeps the refusal after they give the screens away.
Once scheduled, the morning brief already runs without you. After the nod, the bill is for whoever still owns the refusal.