Slack Developers
< Resource Library

AI Agents & Assistants

Tools

AI Agent & Assistant Framework, Bolt for Python

Overview

The new Agents & Assistants feature provides a foundational framework for developers to build customized AI agents within Slack, introducing key events and APIs for interacting with Slack’s new AI container. The AI Agent container enables customers to connect their custom LLMs to meet their organization’s unique needs. Whether it’s automating workflows, answering queries, or providing intelligent insights, this new surface adapts to specific business requirements meeting the needs of mid-market and enterprise customers. Customers can either build their own AI agent or work with Global System Integrators (GSIs) to create enterprise-level AI solutions tailored to their needs.

Key Features

  • New entry point: AI Agents are easily accessible from the top of the Slack web client, allowing users to toggle between multiple agents as needed.
  • Split-view: Users can interact with the AI Agent in a persistent split-view window, enabling them to multitask while engaging with the agent. The split-view remains open as users navigate across Slack.
  • Suggested prompts: AI Agents leverage context to provide dynamic, relevant prompts based on user activity in Slack channels and DMs.
  • Loading state: New shimmer UX and typing status indicators help set user expectations, showing when the agent is processing and working on a response.
  • Updated App Home: The AI Agent’s landing page now includes a Chat tab for ongoing conversations and a History tab making it easy for users to revisit past interactions, creating a more cohesive user experience. Developers can also customize thread titles to summarize key topics.

Customer Benefits

  • Bring Your Own Large Language Model (BYOLLM): Flexibility to use custom LLMs for creating personalized AI Agents.
  • Reusable interface: The AI agent container's intuitive and flexible interface allows seamless integration with custom business workflows and logic.
  • Improved UX: A streamlined, multitasking-friendly interface allows users to interact with AI Agents without disrupting their work.
  • Centralized conversation management: Users can easily track and reference past AI Agent interactions, ensuring better continuity in ongoing projects and discussions.

Technical Implementation

You can find the main developer documentation at https://api.slack.com/docs/apps/ai   App agents and assistants themselves are not an LLM; rather, they give you the tools and interface to best integrate an LLM for use in Slack. Developers who are familiar with creating conversational chatbots in Slack (AI-powered or otherwise) should not have much difficulty taking advantage of the new features that make Slack apps more discoverable, accessible, and message thread-friendly for users. The new surface area is a natural extension of how developers previously integrated bots into Slack, only introducing a few new events, endpoints, and concepts that power the new experience.  

The app assistant flow

  1. Listen for the assistant_thread_started event and inspect for the context
  2. Respond to the assistant_thread_started event with suggested prompts
  3. Listen for the message.im event and set a typing status
  4. Respond to the message.im event with the desired chat response
 

Developer Tools with Assistant support

SDKs
  • Node Slack SDK @slack/web-api version 6.13.0 now includes support for AI Assistants & Agents. Release notes
  • Python Slack SDK Version 3.33.0 now includes support for assistant.threads.* APIs. Release notes
Bolt Framework

Sample Code

Helpful Links

Demo Video

Watch the demo video to see how the AI Agents feature makes Slack the conversational interface for AI.