How to Set Up an AI Slack Bot for IT Support in Under an Hour

An AI bot sitting in your Slack workspace answers the IT questions that otherwise land on whoever runs IT: password reset steps, access requests, printer refusals, VPN questions. Slack ships an official support-agent sample that gets you there in one sitting:

Verified against the Slack CLI and Slack's support-agent sample on 2026-08-26.

Prerequisites

Step 1: Install the Slack CLI and Sign In

  1. Install the CLI: on macOS or Linux, run curl -fsSL https://downloads.slack-edge.com/slack-cli/install.sh | bash. A Windows installer is available from the same docs page.
  2. Run slack version to confirm the install.
  3. Run slack login and complete the authentication prompts for your workspace.

Step 2: Create the Support Agent From Slack's Sample

  1. Run slack create agent.

That single command scaffolds Casey, Slack's IT support agent sample. Out of the box it lets users pick an issue category from the App Home tab, describe problems in a direct message where it keeps context across follow-ups, and answer an @mention from any channel it has joined. It reacts with 👀 while it works an issue and ✅ when the issue is resolved, so the channel can see status at a glance.

Step 3: Choose the Claude Folder and Add Your API Key

  1. Change into the framework folder: cd claude-agent-sdk. The sample carries parallel folders for the OpenAI Agents SDK and Pydantic AI; every folder implements the same agent.
  2. Copy the environment template: cp .env.sample .env.
  3. Open .env and set ANTHROPIC_API_KEY to your key from the Claude Console.

Step 4: Run the App and Install It to Your Workspace

  1. Run slack run.
  2. When prompted, select Create a new app and choose the workspace you logged into.
  3. Wait for the confirmation line: ⚡️ Bolt app is running!

Step 5: Test It in Slack

  1. Open a direct message with the bot and describe a test issue.
  2. Watch for the 👀 reaction, then the threaded reply.
  3. Invite the bot to a public channel and @mention it with a question; it should answer in a thread without leaving the channel.

Step 6: Point It at Your Real IT Systems

The sample answers from mocked data, which is the honest boundary of the first hour: you now have a working bot, and making its answers true for your company is the follow-on project. Swap the mocked tools for your own sources in the agent code:

Keep slack run for development only. For an always-on bot, deploy the app to a host that stays up and install it with a permanent token from OAuth & Permissions.

Verify

Troubleshooting

Would you like an IT support bot answering in seconds, wired to your real ticketing and identity systems, without owning the setup yourself? We can help. Book a free call and we will stand it up as part of your managed IT.

Verified against the Slack CLI and Slack's Casey support-agent sample (Claude Agent SDK) on 2026-08-26. Vendor docs: https://docs.slack.dev/ai/agent-quickstart, https://docs.slack.dev/quickstart.