🧭 Claudeforce Open Beta Is Live: Salesforce CRM Now Reachable Directly from Claude

The Claudeforce open beta is now live for all Salesforce Team and Enterprise customers. Announced August 26 as part of a deepened Anthropic–Salesforce partnership (Salesforce has invested $5 billion in Anthropic and committed to spending $300 million in Claude API tokens in 2026), Claudeforce embeds Salesforce's CRM data, workflows, and business logic into Claude via the Model Context Protocol, allowing Claude to reason over live revenue context and take governed action without requiring users to switch applications.

What Claudeforce ships with today

• 37 prebuilt sales skills covering pipeline updates, deal summaries, forecast rollups, account history lookups, contact enrichment, and Slack follow-up drafting
• Bidirectional data flow: Claude reads from and writes back to Salesforce using the same role-based permissions and field-level security already configured in your org — no separate Claude permissions model to manage
• Agentforce integration: Claude can hand off or receive tasks from Salesforce's Agentforce agents, enabling hybrid flows where Claude handles open-ended reasoning and Agentforce handles structured workflow execution
• Governed action: all CRM writes require a human confirmation step in the beta; fully autonomous writes are on the roadmap for late 2026

What the MCP architecture means in practice

Claudeforce uses MCP to expose Salesforce as a Claude tool server. When you ask Claude "What deals over $200K are at risk this quarter?" it calls the Salesforce MCP endpoint with your live org credentials, retrieves the filtered pipeline view, and synthesises a response — all in the same Claude conversation. Critically, your Salesforce data is not sent to Anthropic for training; it flows only during the inference call, scoped to your session.

Additional prebuilt skills targeting marketing, service, and revenue ops are scheduled to roll out through late 2026. Slack-native Claudeforce (answering Salesforce questions directly in Slack channels) is in limited pilot.

🧭 Claude Design Gets Enterprise Brand Guardrails — Admin Lock, GitHub Import, and a Webinar Today

Claude Design — the visual prototyping tool that launched in research preview on April 17, 2026 — has received a significant Enterprise upgrade. Organisations on Team and Enterprise plans now have admin-level brand controls that constrain what Claude Design can produce, closing the gap between fast AI-generated mockups and approved corporate identity standards.

What's new for Enterprise admins

• Admin-locked design system: a designated admin role can approve one canonical design system (typography, colours, spacing, component rules) and lock edits org-wide — individual users cannot override the approved system
• GitHub import: design systems can be ingested directly from a GitHub repository, a Figma export, or a file upload — no manual copy-paste of tokens
• Brand guardrails on by default: for Enterprise organisations, brand guardrails are enabled by default on activation; for Team, they are opt-in and require an admin to publish the design system before they take effect
• Granular access control: admins can enable Claude Design for specific groups or departments via custom roles, rather than org-wide

How to set it up

Claude Design is disabled for Enterprise organisations until an admin enables it at claude.ai → Organization Settings → Features → Claude Design. Once enabled, go to claude.ai/design → Brand → Import to load your design system. After you publish it, the guardrails apply to every user in scope — Claude Design will refuse to generate outputs that violate the approved typography or colour palette.

🧭 Lutnick at G20: Anthropic Has Done What the Administration Asked, Is "Back on the Right Side"

Speaking at the G20 AI ministers meeting on September 2, 2026, US Commerce Secretary Howard Lutnick said the administration trusts Anthropic, telling reporters that the company is "back on the right side." He introduced Anthropic co-founder Tom Brown to the assembled G20 ministers — a public signal of restored goodwill after months of friction over military AI contracts.

Background: what caused the friction

Earlier in 2026, the US Department of Defense, under Secretary Pete Hegseth, blocked Anthropic from certain military contract programmes after Anthropic declined to allow Claude to be used for domestic surveillance or autonomous weapons deployments. The standoff became public when Anthropic filed suit; a US federal judge subsequently ruled in Anthropic's favour on the scope of the contract restrictions.

What Lutnick's statement means for Anthropic

• Anthropic retains its position in US AI policy discussions, including potential government procurement, export controls, and safety framework consultations
• The G20 ministerial setting — where Anthropic's co-founder was introduced by the Commerce Secretary — suggests Anthropic will be part of the US AI delegation at multilateral forums going forward
• The episode illustrates that Anthropic's refusal to cross specific safety lines (autonomous weapons, domestic surveillance) did not ultimately cost it its US government relationships — the administration recalibrated rather than permanently excluding the company