AI chatbots are only as helpful as the context and data they have access to, and Anthropic is making more partnerships with the most important enterprise and business apps.

On Monday, Salesforce announced support for Anthropic’s Model Content Protocol (MCP) Apps, including the launch of bidirectional extensions in Claude, which bring business data, such as Slack or Agentforce 360, directly into the chatbot for reference and taking action.

MCP is Anthropic's open standard for connecting AI to data systems, and MCP Apps lets companies connect their applications to Claude for direct interaction, not just simple references. For instance, with MCP App capabilities in Slack, users can search their messages and generate, edit, and format drafts in Claude that are sent back to Slack.

“With Anthropic’s new MCP App framework, Slack and Salesforce can provide user-authorized context to Claude in a secure, standardized way, ensuring access is scoped to each user’s profile and permission settings,” said Nick Johnston, SVP of Strategic Tech Partnerships at Salesforce. “This is also how we will deliver Salesforce-generated AI insights within Claude.”

The integration will eventually expand to Agentforce 360, enabling users to leverage Salesforce-native Agentforce actions to perform tasks ideated or recommended by Claude. The idea is to integrate the workflows so users can continue to rely on Agentforce to execute while leveraging the capabilities Claude provides.

“With the planned Agentforce Sales integration with Claude, a sales rep will be able to ask Claude for their highest priority leads, Claude will call Agentforce to generate AI insights grounded in your trusted customer data in Salesforce, and the rep will then get a live, interactive list of leads that can be used to create personalized outreach,” said Johnston.

To address security concerns, the company reassures users that Agentforce apps are built on Salesforce’s existing permissions and security controls, ensuring customer data remains within the platform.