Slack Opens Up Conversational Data to Power Smarter AI Agents

Slack aims to help developers leverage the vast amount of conversational data stored on its platform. The goal: Build better, secure, and context-aware AI agents for the enterprise. To achieve this, the Salesforce-owned firm is launching not only a real-time search API, but also a Model Context Protocol server and other new tools.

“The future of work is undeniably agentic, and the success of AI depends on its seamless integration into human workflows,” Slack Chief Executive Denise Dresser says in a release. “Our latest Slack platform innovations create the secure, data-rich environment necessary for AI agents to become trusted companions.”

AI apps require access to as much data as possible to generate accurate results. In the workplace, this includes both structured and unstructured data stored in systems and third-party apps, as well as team communications. Messaging platforms like Slack contain rich conversational data that AI agents can leverage to work more effectively.

Today’s announcement is another milestone in Slack’s AI policy evolution. In 2024, the company faced backlash from users who were frustrated that Slack would use their data to train “global models” that power its channel and emoji recommendations, as well as search results. Then, in 2025, the company updated its terms of service to ban third-party developers from indexing, copying, or permanently storing Slack messages via its API. Now, it’s opening up those conversations directly to AI agents.

Real-Time Search API

This new data feed provides apps with real-time access to the rich, contextual conversational data inside Slack. When integrated, AI agents will receive the most current discussions, files, and channels in an organization’s instance. Slack asserts that the RTS API comes baked with enterprise security, will always respect existing access permissions, and retrieves data in a “focused, use-case-driven way.”

Remember the previously mentioned terms of service update? Slack states its real-time search API is the primary way for third-party apps to access its data, and even then, there are restrictions.

“Businesses today conduct their most important conversations across multiple Slack channels, creating a wealth of institutional knowledge that hasn’t been accessible to AI assistants like Claude,” Anthropic’s Head of Product for Claude, Scott White, remarks in a statement. “Slack’s real-time search API changes that—Claude can now draw context from your team’s actual discussions to provide answers grounded in your real work, then deliver those insights directly back into your workflow. This integration enables Claude to provide trusted intelligence that’s aligned with how your team actually operates.”

Model Context Protocol Server

The launch of an MCP server makes it easier for AI models, apps, and agents to interact with Slack. Now, there is a universal way for AI to connect to data sources and tools. Developers can use a single, consistent protocol instead of having piecemeal integrations in their applications, and they will no longer need to define every task their agents should perform manually.

This project is the result of three months of work by Slack. In July, the company disclosed its support for MCP. At that time, it also announced that it would develop its own server in collaboration with Anthropic, the creators of the protocol.

With the real-time search API and the MCP server, developers have more flexibility to create contextual and intelligent AI-powered experiences that automate workflows. This means that bots can search, analyze, and act on Slack data, functioning as virtual teammates rather than acting as legacy bots, which are no longer supported.

Other Developer Tools

Alongside RTS and the MCP server, Slack is releasing Slack Work Objects and agentic building tools.

What are Slack Work Objects? It’s a standard dictating how external data, such as detailed information, images, and documents, will be displayed in Slack. It’s creating a unified rich app preview for everything from file descriptions to embedded images. And then, when team members click on these artifacts, they’ll be able to take action right within the Slack workspace, all without context switching.

As for the agentic building tools, Slack is making available prebuilt Block Kit Tables and updated resources, such as the CLI for Bolt apps. All of these upgrades are intended to help developers build and customize Slack for their organizations, streamlining the entire development lifecycle.

When Is All This Available?

Slack shares that the real-time search API and its MCP server are now available in closed beta. It estimates general availability in early 2026. However, third-party AI agents using these capabilities can be found in the Slack Marketplace today. As for Slack Work Objects, that’s expected to launch later this month. The other agentic building tools are available now.

Featured Image: Credit: Ken Yeung