Salesforce opens Slack for external AI

Salesforce wants to expand the instant messaging service for Slack as an AI interface. This was already announced at the end of 2024 with the "Slack AI" add-on and the "AI Workflow Builder" and is now continuing with developer tools. The manufacturer recently announced two interfaces for the Slack developer platform: a Real-Time Search API (RTS API) and an MCP server (Model Context Protocol). This should enable developers to build secure, context-related AI agents directly on the conversation data in Slack.

Until now, the unstructured content of Slack channels has remained largely inaccessible to artificial intelligence. In many companies, they contain discussions, decisions, and feedback from customers. In future, large language models (LLMs) should be able to tap into this resource to provide more relevant answers to customer questions.

According to Salesforce, companies are struggling with fragmented software: jumping between tools, copying, and pasting or context switching reduces productivity. The manufacturer wants to bring Messenger Slack into play as a central point of contact, where a sales employee can update CRM entries via an AI agent without having to switch to another application. An IT colleague can resolve support queries.

Search API and MCP server

The RTS API is designed to provide AI applications with secure access to current messages, files, and channel histories in Slack. According to the manufacturer, it complies with existing authorisations. It should only deliver the desired data fragments in response to requests, instead of mass downloads. The MCP server, the counterpart at protocol level, translates between LLMs and Slack data sources. Developers should thus be able to build generic agents without having to integrate and maintain specific applications.

Salesforce emphasises its security architecture for companies, including granular access rights, audit functions and GDPR-compliant data storage. The manufacturer is also introducing "Slack Work Objects", which integrate external data such as images, text, and tables into the chat in a standardised way. The Agentic Developer Tools are designed to support developers in building Slack applications, AI agents and workflows efficiently and securely, for example with ready-made components for user interfaces.

Partners and third-party providers

Salesforce already has several tools from third-party providers such as Google, Anthropic and Dropbox in Slack: Anthropic is integrating its Claude model into Slack channels, with access to search functions, data source connection and document analysis. Google is expanding access to company knowledge with Agentspace and connecting this with Slack data in both directions. Dropbox integrates the RTS API into its AI-supported search tool Dash. Perplexity combines web search with team conversations for contextualised answers. Notion AI searches conversations in public and private channels via the Slack app and integrates results into the Notion collaboration system.

RTS API and the MCP server are initially only available in a closed beta programme. End customers should be able to access them in early 2026. The Slack Work Objects will be available to all developers at the end of October.

(akr)