Salesforce Lets Claude Invoke Slack and Agentforce Workflows
The enterprise giant lets Anthropic’s chatbot access collaboration and CRM tools, keeping AI useful, trusted, and under control
Salesforce and Anthropic are back with new integrations that make it easier for enterprise workers to use their preferred AI tools on the job. Today’s announcements build on a longstanding partnership between the two tech firms, dating back to 2024. Teams can now use Slack’s Model Context Protocol (MCP) App directly in Claude, transforming the chatbot into a thinking workspace.
In addition, Salesforce is introducing a connection between its Agentforce 360, Data 360, and Commerce 360 apps and Claude.
“Enterprises need more from AI than powerful models,” Nick Johnston, Salesforce’s senior vice president of strategic tech partnerships, says in a statement. “They also need a reliable way for those models to operate inside real business environments. By partnering with Anthropic, we are bringing Salesforce directly into our customers’ flow of work and providing the execution layer with context, data, governance, and trust.”
As it advances its vision of the agentic enterprise, Salesforce may acknowledge that AI can’t live entirely within a walled garden. To gain traction, it needs to show up where customers and workers already are—including the tools they’re using daily. Frameworks such as MCP and A2A make this possible, and the Slack-Claude tie-up offers a good example of how Salesforce is opening its platform without giving up control.
In October, Slack announced the launch of its MCP server that makes it easier for AI models, apps, and agents to interact with the team communication and productivity app. Through text, users could securely search, retrieve, and generate content across Slack conversations. However, today that evolves beyond simple text-based chats, giving users more control over how AI-generated work is structured, reviewed, and shared.
For example, users could ask Claude to prepare a message summarizing the conversation that unfolded in a Slack channel about a specific incident. Claude would then search through Slack to understand what happened before creating a draft message with the appropriate—and editable—summary. A click of a button will then transmit that approved message directly to the relevant Slack channel.
While similar functionality may already be possible using Slack’s new built-in Slackbot, the move is less about duplicating features and more about giving users the flexibility to work with AI tools they’re already comfortable with—without forcing unnecessary context switching. Users will receive responses grounded in Salesforce context, permissions, and real customer data—Claude won’t guess, and there will be minimal hallucinations.
At the same time, Slack’s MCP server provides the underlying security, governance, and trust layer, ensuring enterprise data stays protected and isn’t inadvertently exposed or used for model training.
“By embedding Slack’s MCP App capabilities into Claude, Salesforce makes AI a collaborative assistant rather than an autonomous actor, helping teams move faster without sacrificing trust, accountability, or control,” the company writes in a release.
The news doesn’t stop with Slack—Salesforce is also integrating Anthropic’s chatbot with its Agentforce 360, Data 360, and Customer 360 apps. After receiving a prompt, Claude will call Agentforce, ask it to perform reasoning and execute actions in Salesforce, and receive grounded, trusted, and governed responses. Just as Salesforce positions Slack as the front door to the agentic enterprise, it’s increasingly taking the same view of Claude—not as a system of record, but as a conversational entry point into enterprise workflows that remain firmly under Salesforce’s control.
Why take this approach instead of steering users back to Salesforce’s own apps? Because Salesforce understands it doesn’t need to own the interface to own the outcome. By supporting multiple conversational entry points, the company keeps data, workflows, and activity anchored to its platform, even as users work elsewhere. Salesforce may not be the default UI for every interaction, but it remains the execution engine, collaboration layer, and system of record that ultimately moves work forward.
And while Salesforce has worked closely with Anthropic for years, it isn’t betting on a single model provider. The company maintains similar integrations with OpenAI’s ChatGPT and Google’s Gemini, reflecting a broader strategy that lets customers choose the AI models and conversational interfaces that best fit their needs, while Salesforce continues to provide the underlying data, workflows, and governance.
Thanks for writing this, it clarifies a lot; your insight that AI integration needs to move beyond a walled garden and directly into user workflows is absolutley brilliant and so important for real-world adoption.