Your AI agent in Slack
Connect your Fleece AI Team agent to Slack and it starts answering in channels and direct messages. The agent reads the message, applies the instructions on its profile, and responds — in text, and in its cloned voice when it sends a voice message. It draws on the same knowledge sources, follows the same personality traits, and carries the same identity as on WhatsApp, email or a video meeting. A question in #support and a question on WhatsApp get the same answer because there is one agent behind both.
At a glance
- Provider
- Slack API (OAuth)
- Message types
- Channels, direct messages, threads, voice messages
- Setup
- OAuth connection from the Fleece AI Team dashboard
- Voice
- Same cloned voice as every other channel
- Per-channel routing
- Different agents can be routed to different channels
- Plan
- Included on Team and Scale plans
The channel your team already lives in
Slack is where internal conversation happens — questions about the refund policy, integration questions during a review, status updates that need an answer. Instead of asking a human colleague who may not be at their desk, the team messages the agent.
The agent appears as a Slack user — the one you configured in the Fleece AI Team dashboard. It has the name you gave it and can respond in channels, direct messages and threads. It reads the context of the conversation and responds accordingly.
The agent's knowledge on Slack is the same as everywhere else. It does not have a separate knowledge base for Slack conversations; it reads the same system prompt, the same documents and the same URLs that define it across the board.
Per-channel routing
Different channels can be routed to different agents. A question in #support goes to the support agent, a question in #engineering goes to the engineering agent. The routing is configured in the Slack section of the dashboard.
This is useful when you have multiple agents with different expertise. The support agent knows the refund policy and order status; the engineering agent knows the API documentation and integration details. Routing ensures each question reaches the right agent.
Within a channel, the agent follows its instructions. If someone asks the support agent an engineering question, it responds according to what it knows — and if it does not know, it follows the escalation instructions you gave it.
Threads and context
The agent participates in threads — the Slack feature that keeps related messages grouped together. It reads the thread context before responding, so it can answer follow-up questions and maintain the flow of the conversation.
It also reads the channel history when needed. If a question references a message from earlier in the channel, the agent can read that message and respond to it. This is different from a chatbot that only sees the current message — the agent understands the conversation it is part of.
What the agent cannot do in Slack
The agent cannot access Slack's administration functions — it cannot change workspace settings, manage users, or modify channel permissions. It is a participant, not an admin.
It cannot see private channels it has not been added to. If a channel is private, the agent must be added to it before it can participate. This is a Slack limitation, not a Fleece AI Team one.
The agent does not initiate conversations in Slack — it responds to messages that are sent to it. You can instruct it to post in a channel, but it cannot start a conversation without a trigger.
How to connect your AI agent to Slack
- 01
Create or select the agent
Open the agent's profile in the Fleece AI Team dashboard. The portrait and voice you generate there are what the agent shows in Slack.
- 02
Connect Slack from the Channels page
Go to the Channels section of the dashboard, find Slack, and connect. The platform handles the OAuth flow — you authorise the connection in your Slack workspace.
- 03
Configure channel routing (optional)
If you have multiple agents, configure which agent answers in which channel. By default, all channels are routed to the first connected agent.
- 04
Test with a message
Send a message to the agent in a channel or direct message. It should respond using the instructions and knowledge on its profile.
Questions, answered
Does the agent see all Slack channels?
+
No. The agent only participates in channels it has been added to — or channels that are routed to it in the dashboard. Private channels require the agent to be explicitly added.
Can the agent respond in voice messages on Slack?
+
Yes. When the agent sends a voice message, it uses the cloned voice from its profile — the same voice it uses on phone calls and in video meetings.
Can different agents answer in different channels?
+
Yes. Per-channel routing lets you direct #support to the support agent and #engineering to the engineering agent. Configure it in the Channels section of the dashboard.
Does the agent see Slack file attachments?
+
The agent reads text messages and can extract text from images. It does not process all file types — a PDF or spreadsheet attachment is not opened by the agent.
Can the agent start a conversation in Slack?
+
The agent responds to messages. You can instruct it to post in a channel, but it cannot initiate a conversation without a trigger.
What happens when the agent does not know the answer?
+
It follows the instructions on its profile — it can say it does not know, suggest documentation, or defer to a human colleague in the channel.
Read next
- AI agent on email
The same agent, answering asynchronously — where the conversation continues over hours rather than seconds.
- Send an agent into a video meeting
The agent joins the meeting your team is having — then keeps answering in Slack when the call ends.
- All channels
How one agent keeps the same identity across seven channels — same instructions, same knowledge, same voice.
Connect this channel
Create an agent and connect Slack from the Channels section of the dashboard. One agent, seven channels, one identity.