Dust.tt in Slack: How to Set Up and Call Your Agents

Slack is where most teams want an AI agent to live, because it is where the question gets asked in the first place. Dust supports this in two separate ways, and confusing them is the source of nearly every problem people hit.

The first is the bot: you mention Dust in a channel and an agent replies. The second is Slack as a knowledge source, where Dust indexes the messages in channels you select so agents can answer from them. Setting up one does not set up the other.

This guide covers both, in the order you will actually need them, and ends with the failure list from Dust's own troubleshooting page. That last section is worth reading before you start, because it explains most of what looks like a broken integration.

What do you need to install Dust in Slack?

You need a Slack owner or admin to install the app, and a Dust workspace admin to enable the integration on the Dust side. Dust's admin quickstart describes installing and authorizing the Dust app in Slack, then adding it to the channels where your team should be able to use it.

If you are not sure which Slack role you hold, Dust's troubleshooting page walks through checking it: click your workspace name at the top left, open settings and administration, go to manage members, and find yourself in the list.

• Installing the app: requires a Slack owner or admin role. This is a one-time action for the whole workspace.
• Inviting the bot to a channel: any member of that channel can do it, which means you do not need an admin for every new channel.
• Managing apps afterwards: Dust's docs list owner, admin, or app manager as the roles that can manage installed apps.

How do you get the Dust bot into a channel?

Type the slash invite command with the Dust bot's name in the channel, then press enter. Dust's troubleshooting documentation gives this as the way to add the bot, after which it responds to mentions in that channel.

Private channels are not an exception you can configure around. Their documentation states that Slack apps cannot see private channels until explicitly invited, and describes this as a Slack security behaviour rather than a Dust limitation: private stays private until you grant access.

Getting Dust answering in Slack

1. Install and authorize the Dust app in Slack — A Slack owner or admin does this once. Dust's admin quickstart covers it as part of workspace setup.
2. Enable the Slack integration in Dust — Their troubleshooting page points at workspace management, then workspace settings, then integrations, as the place to verify the toggle.
3. Invite the bot to each channel — Use the slash invite command with the Dust bot name inside the channel. Any channel member can do this, including for private channels.
4. Wait a few minutes for Slack to propagate — Dust's guidance suggests waiting a couple of minutes and refreshing the connection manager rather than assuming it failed.
5. Call the default agent — Mention Dust in a channel or send it a direct message to check the round trip works before you involve anyone else.
6. Call a named agent — Dust's documentation shows the form @dust followed by a plus sign and the agent name, for example a sales helper or a code reviewer you built.

There is a second calling form in their troubleshooting examples, using a tilde instead of a plus before the agent name. If the plus form does not resolve your agent, that is the variant worth trying before you conclude something is broken.

You can also stop typing the agent name every time. Dust's documentation describes linking an agent to a Slack channel through the manage channels option in Slack preferences, so that mentioning Dust in that channel calls your chosen agent instead of the generic one.

Channel linking is the setting that decides whether people actually use this. Asking a colleague to remember a handle works for the person who built the agent and almost nobody else, so link the support agent to the support channel and let the room do the remembering.

Remembering which handle belongs where is the kind of overhead we designed out at Sistava. You hire an AI Employee, and it is a named colleague rather than a handle you have to teach people, so the room already knows who it is talking to.

How do you make Dust read your Slack channels?

Inviting the bot makes a channel visible to Dust, but it does not index it. Their troubleshooting page is explicit that syncing requires separately selecting the channel under spaces, then connections, then Slack, using the add or remove data option.

One consequence is permanent and worth telling your team about upfront. Dust's docs note that Slack's API only provides messages from after the bot joined a channel, so history before the invitation is not retrievable. There is no setting that changes this.

Once the bot is answering and the right channels are indexed, most teams want the next thing: an agent that starts a conversation rather than waiting to be mentioned. That is where Slack workflows come in, and it is the part with real prerequisites.

Can a Dust agent respond inside a Slack workflow?

Yes, but it needs to be whitelisted by Dust first. Their documentation describes building the Slack workflow with a trigger such as a scheduled time or an emoji reaction, then emailing their support address to have that specific workflow allowed.

The request needs your Dust workspace address, the exact names of the spaces involved including any restricted ones the agent must reach, and the exact name of the Slack workflow. Their docs note the whitelisting applies at the workflow level, so you can change which agent it calls afterwards without asking again.

There are visibility rules on the Slack side too. Dust's documentation says it can only interact with workflows it has permission to access, that the workflow should be available to everyone internal or include Dust in the relevant channels, and that workflows restricted to managers only are out of reach.

Why is the Dust bot not responding?

Nearly always one of four things, and Dust publishes the list. Work through it in order before assuming the integration is broken, because the fixes are quick and the causes are mundane.

Their docs also flag scale as a factor: on workspaces with more than a thousand channels, the channel list can lag behind reality for a few minutes. Refreshing after a short wait is the documented answer rather than repeating the setup.

Taken together, the Slack setup is not hard, it is just a chain. App installed, integration enabled, bot invited per channel, channel selected for sync, agent linked to the channel. Every link is small and every one of them can silently be the missing piece.

Chains like that are what we removed at Sistava. You hire an AI Employee, connect Slack once, and tell it in plain English which conversations it should look after. There is no per-channel wiring to keep in your head and nothing to re-link when a new channel appears.

If you are staying with Dust, the habit that pays off is to prove each link before adding the next. Get one channel answering, index that same channel, then link an agent to it. A working chain in one room beats a half-configured one across twenty.

Sources

• Dust documentation, Slack troubleshooting
• Dust documentation, Slack workflows
• Dust documentation, Admin quickstart
• Dust documentation, Managing Agents
• Dust documentation, What data do the agents have access to?
• Dust documentation, Connections

Slack is the right home for this kind of tool, because the question and the answer end up in the same place your team already searches. Get the chain right once and it mostly runs itself after that.