Slack dev tools pave the way for new AI agents

October 1, 2025 Yanac

A new wave of AI agents are coming to Slack, enabled by the addition of several developer tools.

It’s now easier to build and launch agents that tap into real-time conversation data held in the collaboration app, Slack said on Wednesday, and for users to interact with third-party AI assistants such as Claude or Perplexity directly from their Slack workspace.

The result is scalable agent integrations that can provide users with more accurate and relevant information in the flow of their work, Slack said.

“For employees, this means agents that actually help, right next to where you talk to your colleagues: prepping talking points, analyzing docs, managing incidents, tracking projects — all without the constant app-switching that kills productivity,” said Rob Seaman, Slack’s chief product officer, in an email to Computerworld. “For IT, it’s a cleaner path to deploy agents at scale while keeping data locked down with the right permissions.”

[ Related: Agentic AI – Ongoing news and insights ]

Among the additions are a new real-time search API and a model context protocol (MCP) server. Both are aimed at third-party vendors, as well as organizations that want to build their own custom AI agents.

The real-time search (RTS) API grants agents access to converstation data and files in Slack channels without the need to bulk download the data or store it elsewhere. This is done while respecting existing access permissions, Slack said, and will only return messages or files relevant to a query.

In Notion, for instance, it’s possible to search and surface content from Slack public and private channels, as well as direct messages, in line with a user’s access permissions. Similarly, Dropbox Dash — Dropbox’s enterprise search tool — can also highlight the latest conversations occurring in Slack.

The Slack MCP server is based on an open-source framework developed by Anthropic that standardizes how agents access information and execute tasks. This reduces complexity for developers when connecting agents to Slack.

“Fragmented integrations are replaced with a single, consistent protocol that adheres to the user’s specific permissions,” Slack said in a blog post. “Developers don’t have to manually define every possible task an agent can perform or implement complex, service-specific integrations for each LLM.”

Anthropic and Perplexity make use of Slack’s MCP server to connect their agents to the collaboration app.

With Perplexity’s AI assistant, the MCP server makes it easier to ground queries in the context of Slack conversations, resulting in answers that are more accurate and relevant. “This broadens the use cases users can tackle with Perplexity and the confidence they feel in the results they get,” said Frank te Pas, head of enterprise product at Perplexity, in a prewritten statement.

Perplexity uses Slack’s new MCP server to access data from Slack conversations.
Slack

Both the RTS API and MCP server are available in a closed beta currently, ahead of general availability in early 2026. However, the third-party agents are available to Slack customers via the Slack Marketplace from today. Pricing of individual AI apps and agents will be determined by the vendor.

Slack also announced Work Objects on Wednesday, a tool that helps developers connect apps to Slack conversation data with “rich app previews” that include information such as file descriptions and embedded images. Work Objects standardizes how third-party data is displayed in Slack, and lets users take actions, such as marking Asana tasks as complete, without switching screens.

Finally, Slack has created new agentic developer tools that it said will streamline the development lifecycle, with “AI best practices, prebuilt Block Kit Tables, and updated resources that includes the CLI for Bolt interface.”

Slack Work Objects will be generally available to all developers in late October. The new agentic developer tools are available today.