What is Claudeforce? Salesforce and Anthropic bring Claude to your CRM

Claudeforce is the expanded partnership between Salesforce and Anthropic announced on 26 August 2026, presented as “the #1 AI meets the #1 AI CRM”. The core: Claude becomes more deeply woven into the Salesforce platform, and Salesforce in turn becomes usable from within Claude itself. For organisations running on Salesforce this is more than a press release: it confirms that the AI agent is becoming a full entry point to your CRM, alongside the browser and the mobile app. At Outbirds we have been building with Claude and MCP on Salesforce for a while; in this article we explain what Claudeforce covers, what becomes available when, and what you should already be doing.

What exactly is Claudeforce?

Claudeforce is not a separate product but an umbrella for three tracks in which Salesforce and Anthropic deploy each other’s technology by default.

Track one: Claude in Salesforce. Claude becomes a default reasoning model within Agentforce, including the Atlas Reasoning Engine, Agentforce Vibes, Agentforce Coworker and Agent Builder. Important for regulated sectors: Claude runs through Amazon Bedrock within the Salesforce Trust Boundary, so data does not leave Salesforce’s security perimeter. There is no forced single-model architecture; other models remain available in Agent Builder.

Track two: Salesforce in Claude. A plugin that gives Claude direct, governed access to your Salesforce org, launching with 37 prebuilt sales skills for meeting preparation, deal health reviews and pipeline analyses, among others. Sales reps can reason over their live revenue context from within Claude and have approved actions executed directly, with all the org’s business rules fully in force.

Track three: Claude in Slack. Claude becomes the default model behind Slackbot, Claude Tag and Slack Code, bringing the same intelligence to the place where teams already collaborate.

How does Salesforce in Claude work under the hood?

Salesforce in Claude runs on what Salesforce calls AIforce: the enterprise layer that securely exposes data, workflows and business logic to AI agents through MCP servers, APIs and CLI tools. This builds directly on Headless 360, the TrailblazerDX programme (April 2026) that exposed every platform capability as an API and MCP tool. If you read our earlier explainer on headless Salesforce, you will recognise the pattern: the agent is the new consumer of the platform.

The security model is the most important news for administrators. An administrator connects the org once; after that, every user works through their own OAuth consent, and every action runs under that specific user’s permissions. Sharing rules, field-level security, validation rules, flows and triggers apply in full: a write from Claude is a normal Salesforce write, not a backdoor. For team channels through Slack, Salesforce works with a dedicated integration user with scoped access.

When is Claudeforce available and what does it cost?

The Salesforce in Claude plugin has been running as a pilot with selected customers since the announcement; the open beta is planned for September 2026, with additional skills later in 2026. Claude as a model within Agentforce and the Slack integrations follow Salesforce’s regular release rhythm. Technically, API version 67.0 or higher is the foundation for MCP access.

Little is official on pricing yet: the expectation is consumption-based pricing, tied to API usage and model inference, on top of your existing licences. So expect governance over agent usage to become a cost question too: an agent firing queries without restraint costs money and API budget.

What does Claudeforce mean for your organisation?

For service-intensive and asset-intensive organisations, the scenario we have been building for a while becomes concrete: a service manager asking Claude which work orders are at risk this week, a field service planner having a reschedule calculated, or an account manager pulling up the installed base and open cases straight from a customer conversation. The agent as the entry point, Salesforce as the system of record.

For nonprofits this connects to the line Salesforce started at the end of 2025 with Agentforce Nonprofit: AI agents for fundraising, programme management and donor support. Claudeforce puts frontier intelligence underneath, within the same governance. How we deploy Salesforce for the sector is on Salesforce for nonprofits.

And for every organisation: the first 37 skills are sales-focused, but the architecture underneath (AIforce, MCP, per-user permissions) is generic. Skills for service, field service and other clouds will follow; whoever gets their org foundation in order now will reap the benefits immediately.

How do you prepare your org for Claudeforce?

Our concrete advice, in order:

1. Clean up your permission model. Every Claudeforce action runs under the user’s permissions. Overly broad profiles and outdated permission sets become directly visible in what an agent can do. Least privilege first.
2. Get your API foundation ready. Check API versions (67.0+), integration users and your monitoring of API usage and limits.
3. Start with read scenarios. Meeting prep, pipeline insight and deal health are valuable and low-risk. Writes follow once trust and governance are in place.
4. Treat the beta as a beta. Do not hang critical quarter-end processes on functionality still in pilot or open beta.
5. Build experience with agentic working. Organisations already working with AI coding agents and MCP on Salesforce will have a head start in governance, patterns and team skills. That is exactly our daily practice; see agentic engineering.

Through all of this, Salesforce remains the core and the single source of truth; Claudeforce adds a new, intelligent entry point. That is the same movement we are already making with headless and agentic engineering, now endorsed by Salesforce and Anthropic themselves.

Frequently asked questions

What is Claudeforce in one sentence?

Claudeforce is the partnership between Salesforce and Anthropic (announced on 26 August 2026) that makes Claude available by default within Agentforce and Slack, and makes Salesforce usable from Claude through a plugin with 37 sales skills.

Does Claudeforce replace Agentforce?

No. Agentforce remains Salesforce’s agent platform; Claude becomes a default reasoning model within it. Salesforce in Claude is an additional entry point to your org, not a replacement of the platform.

Is my Salesforce data safe with Claudeforce?

The design aims for exactly that: Claude runs within Agentforce through Amazon Bedrock inside the Salesforce Trust Boundary, and actions from Claude run under the individual user’s permissions, with all existing sharing rules, field-level security and validation rules in force. As always: the configuration of your permission model determines the practice.

When can I use Salesforce in Claude?

The pilot has been running with selected customers since late August 2026; the open beta is planned for September 2026, with additional skills later in 2026.

What does Claudeforce cost?

Official pricing has not been announced at the time of writing. The expectation is consumption-based pricing based on API usage and model inference, on top of existing licences.

Does Claudeforce also work for service and field service?

The first 37 skills focus on sales, but the underlying architecture is generic and Salesforce has announced skill expansion for later in 2026. For service- and asset-intensive scenarios we already build comparable agent access today through MCP and Headless 360.